The association between dietary patterns, diabetes and depression

Author's Accepted Manuscript

The association between dietary patterns, diabetes and depression Joanna Dipnall, Julie A Pasco, Denny Meyer, Michael Berk, Lana J Williams, Seetal Dodd, Felice N Jacka

www.elsevier.com/locate/jad

PII: DOI: Reference:

S0165-0327(14)00740-X http://dx.doi.org/10.1016/j.jad.2014.11.030 JAD7121

To appear in:

Journal of Affective Disorders

Received date: 26 September 2014 Revised date: 16 November 2014 Accepted date: 17 November 2014 Cite this article as: Joanna Dipnall, Julie A Pasco, Denny Meyer, Michael Berk, Lana J Williams, Seetal Dodd, Felice N Jacka, The association between dietary patterns, diabetes and depression, Journal of Affective Disorders, http://dx.doi.org/ 10.1016/j.jad.2014.11.030 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

1 The association between dietary patterns, diabetes and depression.

Joanna Dipnall1,2, Julie A Pasco1,3, Denny Meyer2, Michael Berk1,4,5,6, Lana J Williams1, Seetal Dodd1,4,6, Felice N Jacka1,4,7,8. 1

IMPACT Strategic Research Centre, School of Medicine, Deakin University, Geelong, VIC, Australia

2

Department of Statistics, Data Science and Epidemiology, Swinburne University of Technology,

Hawthorn, VIC, Australia 3

NorthWest Academic Centre, Department of Medicine, The University of Melbourne, Melbourne, VIC,

Australia 4

Department of Psychiatry, The University of Melbourne, Melbourne, Australia

5

Florey Institute of Neuroscience and Mental Health, Parkville, VIC, Australia

6

Orygen, The National Centre of Excellence in Youth Mental Health, Parkville, VIC, Australia

7

Centre for Adolescent Health, Murdoch Children’s Research Institute, Melbourne, VIC, Australia

8

Black Dog Institute, Sydney, NSW, Australia

[email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected]

2 ABSTRACT Background: Type 2 diabetes and depression are commonly comorbid high-prevalence chronic disorders. Diet is a key diabetes risk factor and recent research has highlighted the relevance of diet as a possible risk for factor common mental disorders. This study aimed to investigate the interrelationship between dietary patterns, diabetes and depression. Methods: Data were integrated from the National Health and Nutrition Examination Study (2009-2010) for adults aged 18+ (n=4,588, Mean age=43yr). Depressive symptoms were measured by the Patient Health Questionnaire-9 and diabetes status determined via self-report, usage of diabetic medication and/or fasting glucose levels 126 mg/dL and a glycated hemoglobin level 6.5% (48 mmol/mol). A 24-hour dietary recall interview was given to determine intakes. Multiple logistic regression was employed, with depression the outcome, and dietary patterns and diabetes the predictors. Covariates included gender, age, marital status, education, race, adult food insecurity level, ratio of family income to poverty, and serum Creactive protein. Results: Exploratory factor analysis revealed five dietary patterns (healthy; unhealthy; sweets; “Mexican” style; breakfast) explaining 39.8% of the total variance. The healthy dietary pattern was associated with reduced odds of depression for those with diabetes (OR 0.68, 95% CI [0.52, 0.88], p=0.006) and those without diabetes (OR 0.79, 95% CI [0.64, 0.97], p=0.029) (interaction p=0.048). The relationship between the sweets dietary pattern and depression was fully explained by diabetes status. Conclusion: In this study, a healthy dietary pattern was associated with a reduced likelihood of depressive symptoms, especially for those with Type 2 diabetes.

3 Keywords Diabetes, depression, diet, dietary patterns, psychiatry, nutrition.

BACKGROUND Diabetes is a chronic disease with serious complications, affecting approximately 347 million people worldwide (Danaei et al., 2011) and 29.1 million children and adults in the United States (US) (9.3% of the US population) (Centers for Disease Control and Prevention, 2014). Type 2 diabetes is the most common type, with risks related to lack of regular physical activity, unhealthy eating and excess weight. Throughout the western world, obesity and prevalence of Type 2 diabetes is rising, with substantial associated costs to individuals and society. According to the Center for Disease Control and Prevention, one in 10 U.S. adults has diabetes in 2014, with the number of new diabetes cases each year predicted to increase 87% from 8 per 1,000 people in 2008, to 15 per 1,000 in 2050 (Boyle et al., 2010). Diabetes is highly comorbid with major depression (Anderson et al., 2001; Egede and Ellis, 2010) compounding health care use and expenditure (Egede et al., 2002). Depressive disorders now rank second in terms of global disability burden and are expected to be the number one health concern in both developed and developing nations by 2020 (Lim et al., 2013; Mathers and Loncar, 2006; Murray and Lopez, 2013). Sufferers of both Type 2 diabetes and depression are at greater risk for complications over five years (Lin et al., 2010), including mortality (Katon et al., 2005; Roglic and Unwin, 2010), compared to those without depression. Given the bidirectional relationship (Mezuk et al., 2008) and comorbidity between depression and diabetes, it is not surprising that many of the same poor lifestyle factors associated with Type 2 diabetes, specifically diet and physical activity, are also associated with depression (Strine et al., 2008; Weber et al., 2000). Diet is a major lifestyle factor implicated in, and used to manage Type 2 diabetes, and has also been shown to be associated with risk for depression. A systematic review and meta-analysis of the relationship between dietary patterns and depression (Lai et al., 2014) highlighted the variety of dietary patterns investigated with respect to depression, such as healthy (e.g. Mediterranean, Japanese), and unhealthy (e.g.

4 Western), and confirmed that a healthy dietary pattern is associated with a reduced prevalence of depression. Healthy diets such as the Mediterranean diet (Psaltopoulou et al., 2013; Rienks et al., 2012; Sánchez-Villegas et al., 2009; Sánchez-Villegas et al., 2013); a diet high in vegetables, meat, poultry, and dairy (Meyer et al., 2013); a diet high in fruit & vegetables (Kronish et al., 2012), 2012); the Japanese diet (fruit, vegetables, green tea, soya) (Nanri et al., 2010); and a ‘traditional’ diet of vegetables, fruit, fish and unprocessed meat (Jacka et al., 2011; Jacka et al., 2010) have all been shown to have an inverse relationship with depression. A meta-analysis examining the association between adherence to a healthy style Mediterranean diet and the risk of stroke, depression, cognitive impairment, and Parkinson disease (Psaltopoulou et al., 2013) also concluded that adherence to a Mediterranean diet was associated with a reduced risk for these brain disorders. At the same time, an unhealthy style diet high in processed foods (sweets, fried food, processed meats, refined grains, high fat diary)(Akbaraly et al., 2009); ‘western’ (Chocano-Bedoya et al., 2013; Jacka et al., 2010; Nanri et al., 2010); meat & processed (Rienks et al., 2012); and biscuits & snacking (Samieri et al., 2008) have all previously been found to be associated with increased odds for depression. While these studies and reviews were based on observational data, the possible preventive utility of a Mediterranean diet on depression risk was recently assessed in a randomised controlled trial. Using data from a large, multicentre, randomized, primary prevention field trial of cardiovascular disease (PREDIMED), community-dwellers aged 55 years and over and at high risk of cardiovascular disease were randomised to one of three dietary interventions, followed for at least three years. The results demonstrated that the Mediterranean diet combined with nuts reduced the risk for depression, particularly in people with Type 2 diabetes (Sánchez-Villegas et al., 2013). In addition, the Japanese diet has also been found to play a complex role in the development of Type 2 diabetes, where an increased risk was associated with greater rice intake among women and among physically inactive men and decreased risk was associated with greater fish/seafood intake amongst men, and depression risk in Japanese men and women decreased with higher intake of vegetables, fruit, mushrooms, and soy products (Nanri, 2013). As such, the aim of this research was to investigate the interrelationship between dietary patterns, diabetes and depression using a population-based representative sample of US citizens.

5 METHODS Cross-sectional, population-based data from the National Health and Nutrition Examination Surveys (NHANES) (2009-2010) (Centers for Disease Control and Prevention National Center for Health Statistics, 2013) were utilized for this research study. Approximately 5,000 non-institutionalized US civilians aged 20 to 75 per annum responded to a detailed home based interview and physical examinations at mobile examination centres across the country. The sampling methodology involved a stratified, multistage probability process, with data collected annually but released in blocks of two years. Relevant NHANES data files were downloaded from the website and integrated using the Data Integration Protocol In TenSteps (DIPIT) (Dipnall et al., 2014). Across the two years studied, markers of depressive symptoms were available for 4,656 participants, and 4,588 of these were identified with or without diabetes and thus were included in this study. NHANES received approval from the National Center for Health Statistics research ethics review board and informed consent was obtained for all participants.

MEASURES Depression The Patient Health Questonnaire-9 (PHQ-9) (Kroenke and Spitzer, 2002; Kroenke et al., 2001; Martin et al., 2006) was used to assess depressive symptoms. The PHQ-9 comprises depression modules taken from the larger PRIME-MD Patient Health Questionnaire (Spitzer et al., 1999). The nine items used incorporate key depressive disorder diagnosis criteria from the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) (American Psychiatric Association, 2000; Kroenke and Spitzer, 2002; Kroenke et al., 2001). There are a variety of valid measurement tools for depression and depressive symptoms (Holt et al., 2014), one of the most commonly used being the PHQ-9. This is a well-validated measure for detecting and monitoring depressive symptoms (Kroenke et al., 2010) and has sufficient concordance with clinical diagnosis of major depressive disorder (MDD) (Eaton et al., 2007). The nine items consisted of a 4-point scale indicating the degree of severity from 0 (not at all) to 3 (nearly every day). These items were

6 then summed to form a total severity score ranging from 0 to 27. A dichotomous identifier was then ascribed whereby those with a total score of 10 or more were considered depressed (i.e. moderately to severely depressed (Kroenke et al., 2001) ). Diabetes Participants were identified as having Type 2 diabetes if they (1) reported that either a doctor or health professional had told them they had diabetes or sugar diabetes (other than during pregnancy); (2) consistent with Mezuk (Mezuk et al., 2013), reported the use of hyperglycaemic agents, with the medication bottle seen by the interviewer; and/or (3) diabetes based on their fasting blood glucose and glycated haemoglobin tests. In 1997, the diagnosis of diabetes was re-defined by the American Diabetes Association (ADA) and the World Health Organization (WHO) as a fasting plasma glucose level of 126 mg per decilitre (7.0 mmol/L) as this is the level at which a unique micro vascular complication of diabetes, retinopathy, becomes detectable (Alberti, 1998; American Diabetes Association, 2010; World Health Organisation (WHO), 1999). Since this test only reflects glycemia at the moment the sample was taken, there are some disadvantages relying just on this test that may bias the results (Inzucchi, 2012). For this reason the second blood test that measured glycohemoglobin (A1C test), or hemoglobin A1c, HbA1c test, was also used. The A1C test is a blood test that provides information about average levels of blood sugar (or glucose), over the past 3 months (Nathan et al., 2008) and has been primarily test used for diabetes management and diabetes research (Patel and Macerollo, 2010) and may serve as a better biochemical marker of diabetes and more reliable measure of chronic glycemic levels (Committee, 2009). The test is a marker of long-term glycemia but has some disadvantages, predominantly relating to its reliability in people with such illnesses as anemia, after recent transfusion renal disease, racial and ethnic differences. However, for best results only those with the diagnostic cutoff point for diabetes is a fasting plasma glucose level of 126 mg per deciliter (7.0 mmol/L) and more or a glycated hemoglobin level of 6.5% or more (i.e.  48 mmol/mol) were considered to have diabetes.

7

Diet The nutritional assessment component of NHANES included a 24-hour dietary recall interview for participants of all ages, conducted in person by trained dietary interviewers (Beaton et al., 1979; Posner et al., 1982; Thompson and Byers, 1994). Respondents reported the volume and dimensions of the food items consumed in order to estimate portion sizes. A second dietary interview for all participants who complete the in-person recall was collected by telephone approximately 3 to 10 days later. These dietary interviews yielded information on the regular consumption of 25 items (Table 1); these data were used to determine dietary patterns. [TABLE 1]

Covariates Initial covariates included both categorical and continuous variables. Gender, age group, marital status, education, race, smoking status, a level of individual food security, and body mass index (BMI, kg/m²) were grouped into categories. The ratio of family income to poverty and a measure of C-reactive protein (CRP, mg/dL) were retained as continuous variables. Marital status was collapsed into three groups: never married, married/living with a partner and widowed/divorced/separated. Race was collapsed into four groups: Mexican American and other Hispanic, Non-Hispanic White, Non-Hispanic Black, and other. Smoking status was categorized into those who have never smoked, former smokers, and current smokers. Education was collapsed into three groups: grades 11 and below, high school / General Educational Development equivalent, and some college / Associates degree (AA) / college or above. Food insecurity refers to limited or uncertain access to food resulting from inadequate financial resources. Given the established association between food insecurity and obesity (Wilde and Peterman, 2006) and diabetes (Berkowitz et al., 2013; Seligman et al., 2007), the individual adult food security measure from the NHANES study was included as a covariate. Individual-level food security items were administered to all adults 16 years and over in the households that affirmed any Food

8 Security Survey Module (FSSM) item during the household interview after the 24-hour dietary recall in the Mobile Examination Center (MEC). Affirmative responses were tallied for the 10 food security questions and categorized into four groups: full food security (no affirmative response in any of the 10 items); marginal food security (1 to 2 affirmative responses); adult low food security (3 to 5 affirmative responses); and adult very low food security (6 to 10 affirmative responses). All survey participants were eligible for the body measurement component and were collected in the MEC by trained health technicians. BMI was calculated from the respondent’s weight (kg) and standing height (cm) using the standard formula (kg/m²) and grouped into four categories: underweight (BMI under

18.5); normal (BMI between 18.5 to under 25); overweight (between 25 to under 30); and obese (BMI 30 or above). The weighted information for each covariate and its significance to depression is presented in Table 2. [TABLE 2]

STATISTICAL ANALYSIS Initial dietary results were analysed using exploratory factor analysis (EFA), with correlations matrix as input. Factorability of R was examined using four key measures (Tabachnick and Fidell, 2012): 1) Sample size; 2) Inspection of item correlations; 3) Kaiser-Meyer-Olkin (KMO) test of sampling adequacy; 4) Bartlett’s test of sphericity. Factor analysis was deemed appropriate as the unweighted sample size was large, with a ratio of 179 observations per item. There were also several correlations above 0.3, the KMO was greater than 0.6, and Bartlett's test of sphericity was statistically significant (p<0.001, = 0.05). Kaiser’s rule (retaining factors with eigenvalues >1), Cattell’s scree test (Cattell; Cattell, 1965, 1988), parallel analysis (Horn, 1965), and minimum average partial (MAP) correlation tests (Velicer, 1976; Velicer et al., 2000; Velicer et al., 1982; Watkins, 2005) were used to assess dimensionality of the item set. Parallel analysis and MAP tests were carried out using a computer program developed by O’Connor (O’connor, 2000).

9 Following recommendations of Pett et al (Pett et al., 2003), initial factor solution utilised principal components extraction, with an oblique direct oblimin rotation. An oblique rotation was chosen as it was expected that different diets would be correlated. Interpretability of solutions retaining different number of factors was assessed using pattern coefficients, with loadings >0.3 interpreted as statistically significant. Solutions retaining a different number of factors were assessed on the following criteria: conceptual meaningfulness of factors; simple structure (Thurstone, 1947); factor reliability (at least 3, preferably 4, variables defining each factor); a reasonable amount of the total variance explained; and communalities a minimum of 0.2. Examination of assumptions of EFA revealed no contraindications for factor analysis. The sample used was in excess of a minimum sample of 200 recommended for EFA (MacCallum et al., 1996; MacCallum et al., 1999). The determinant was >0.00001 (0.164), indicating that there were no problems with collinearity (Field, 2009). Visual inspection of scatterplots for pairs of randomly chosen items showed no evidence of non-linearity, and there were no univariate outliers. All items were measured on a metric scale with a number positively skewed but no transformations were performed on the items. A multiple binary logistic regression analysis (Dupont and Dupont, 2009; Hilbe, 2009; Hosmer Jr et al., 2013; Long and Freese, 2006) was performed, where the logit is the natural logarithm of the odds of depression, taking the range of values potentially between -λ and λǤ , in order to investigate the relationship of the different diets with depression, taking into account the multistage survey nature of the NHANES data for point estimation, model fitting and variance estimation (Heeringa et al., 2010; Wolter, 2007). Further testing was performed on the final model to take into consideration possible confounding, mediation and moderation. All statistical procedures were performed using Stata V13 software (StataCorp., 2013).

RESULTS Exploratory Factor Analysis for the Diet Factors Visual inspection of correlations matrix identified several correlations more than 0.3 (Appendix 1). For the 26 items, KMO was 0.70 and Bartlett’s test was significant (² (300) = 317378113.4, p<0.001), supporting

10 factorability of the 25 item set. Results of dimensionality tests disagreed on the number of factors potentially underlying the dietary item set: Kaiser’s criterion identified 8 factors with eigenvalues above 1, scree test suggested the presence of 3, 5 or 8 factors (Figure 1), parallel analysis suggested 7 factors, and MAP test suggested 2 factors. The initial 8 factors retained explained 47% of the total variance in the data. [FIGURE 1] A systematic approach was used to solution refinement, with maximum likelihood and principal factor extractions methods evaluated and the optimal solution was a 5 factor solution with four items removed one step at a time (popcorn, sweetened tea/coffee, fruits/sports/energy drinks, and then finally 100% fruit juice). Varimax rotation was used with the final solution as it yielded the strongest loadings on each of the final 5 factors with the most meaningful interpretation. The final 5 factor solution using PCA extraction with varimax rotation explained 39.8% of the total variance. Each of the 5 rotated diet factors explained 9.7%, 8.8%, 7.7%, 7.0% and 6.6% of the total variance. All communalities were above 0.2 for the final 21 items in the analysis. The final solution was close to a simple structure as there were no significant cross loadings across any of the 5 factors. With the exception of factor 5, there were at least three food items that significantly loaded on each factor. There were only two food items significantly loading on factor 5, but the two loadings were each strong enough, and the food items made good theoretical sense to justify the inclusion of this factor. The items significantly loading on the final 5 diet factors are summarised in Table 3, along with the communalities on each food item. [TABLE 3] Factor 1, a Healthy Diet, was characterised by vegetables, leafy/lettuce salad, fruit, cooked whole grains, and whole grain bread. Factor 2, an Unhealthy Diet, was characterised by the consumption of fried potatoes, cheese, red meat, processed meats, pizza, non-fried potatoes, and regular soft drinks. Factor 3, a Sweets Diet, was characterised by the consumption of cookies/cake, chocolate or candy, ice cream, and pastries foods. Factor 4, a “Mexican” Style Diet, was made up of beans, tomato-based salsa, and tomato sauce. Factor 5, a Breakfast Diet, was made up of cold or hot breakfast and milk. Higher scores on these

11 dietary patterns represented increased consumption of these types of foods. After removal of outliers on the five diet factors the scales were transformed into a standardized variable (μ=0, =1) for use in the multiple binary logistic regression analysis.

Binary Logistic Regression Bivariate logistic regression was performed initially and revealed that only two of the five diet factors had a significant bivariate relationship with depression: the healthy diet and the sweets diet (Table 4). [TABLE 4] Initially, simple multivariate models were tested with the inclusion of each of the two diet factors separately to investigate the relationship between diabetes and depression. The initial relationship observed between the sweets diet factor and depression was fully explained by the relationship between diabetes and depression. Once a person’s diabetes status was included in the simple bivariate logistic regression model, the sweets diet factor was no longer significant OR 1.11, 95% CI [0.93, 1.32], p = 0.221) as a person’s diabetes status explained the major proportion of the relationship between the sweets diet factor and depression (diabetes OR 2.17, 95% CI [1.44, 3.29], p = 0.001). Although the healthy diet factor remained a significant predictor of depression with the inclusion of diabetes status (OR 0.70, 95% CI [0.60, 0.83], p<0.001), along with diabetes (OR 2.14, 95% CI [1.40, 3.28], p = 0.002), significant interaction was found between the healthy diet factor and diabetes on depression (OR 0.78, 95% CI [0.61, 0.99], p = 0.045). The relationship between the healthy factor diet and depression significantly differed according to diabetes status. For the final multiple binary logistic model, after controlling for covariates, the significant interaction between the healthy diet and diabetes status remained (Table 5). After controlling for a person’s sweets diet factor score, gender, age, marital status, education, race, smoking status, level of food security, Ratio of family income to poverty, and serum CRP, the relationship between the healthy diet factor score and depression differed according to a person’s diabetes status. BMI status was excluded from final model due to its negative influence on the goodness of fit test for the fixed effects model, with its exclusion not

12 influencing the relationship of the diets to depression and the significance of the interaction between the healthy diet factor and diabetes. The final model fit was acceptable, yielding an insignificant test for goodness of fit for depression (F(9,8)=0.800, p=0.627)(Archer and Lemeshow, 2006). [TABLE 5] Of interest is the difference in the impact of the two diet factors on the probability of depression for each gender is shown in Figure 3. The more a person eats a healthy diet the lower the probability of depression, irrespective of gender. However, the sweets diet has an opposite effect on the probability of depression: the more a person eats a sweets rich diet the higher the probability of depression. [FIGURE 2] The results suggest that for those with diabetes, a healthy diet dramatically reduces their probability of depression compared to those without diabetes (Table 5). This is not the case with the sweets diet, where the probability of depression for those with diabetes is consistently higher than for those without the condition (Figure 3). [FIGURE 3] A sensitivity analysis was performed, where the final model was run across both diabetes subpopulation status (Table 6). The only significant predictors of depression for those with diabetes were the healthy diet factor, education (High school/GED Equivalent; Some College/AA / College or Above) and ratio of family income to poverty. For those without diabetes the significant predictors of depression were the healthy diet factor, gender (Female), age (45 to 55 years), smoking status (Current smoker), and adult food insecurity (Low food security; Very low food security). [TABLE 6] Thus, the impact of the healthy diet factor was found to be strongest in those with diabetes than those without the condition. This result further confirms the potential impact of the healthy diet factor for sufferers of diabetes on their relationship between their disease and the probability of depression.

13

DISCUSSION This study supports previous research in showing that a healthy diet is associated with a reduced probability of depressive symptoms or depression, particularly in those with type 2 diabetes (Lai et al., 2014; Psaltopoulou et al., 2013; Sánchez-Villegas et al., 2013). However, it extends the extant literature in suggesting that one important mechanism linking diabetes and depression is diet quality. This finding is consistent with the results of the recent PREDIMED study, which demonstrated the benefit of a Mediterranean diet with nuts on the risk for depression in people with Type 2 diabetes (Sánchez-Villegas et al., 2013). Our results also support those from another recent dietary intervention in which incident episodes of major depression over two years of follow-up were prevented in older adults with elevated depressive symptoms (Reynolds et al., 2014). In this study, it is notable that the mean BMI of participants was greater than 30, although no data were presented on participants’ diabetes status. Finally, our results are concordant with the findings of recent meta-analyses confirming an inverse relationship between various types of healthy diet and depression (Lai et al., 2014; Psaltopoulou et al., 2013). We also observed that the initial positive relationship between the sweets diet factor and depression was attenuated by the introduction of diabetes as a covariate; this warrants some discussion. As this study utilised cross-sectional data, it is not possible to determine the temporality of the relationships between these three factors. One possibility is that diabetes is on the causal pathway between a diet high in sugar and depression. In other words, high sugar diets predispose to diabetes, which in turn prompts the development of depression. Another possibility is that a high sugar diet is an explanatory factor in the diabetes-depression relationship. There are observational data to show that unhealthy, ‘western’ diets are risk factors for depression (Akbaraly et al., 2009; Jacka et al., 2014; Jacka et al., 2011) and the noxious impact of western dietary components (i.e. saturated fat and refined carbohydrates) on the brain, as well as obesity, is also well described (Kanoski and Davidson, 2011). As such, it may be the case that an unhealthy diet is a key risk factor for both diabetes and depression. There are multiple potential pathways that may explain why a healthy diet may influence the relationship between depression in those with diabetes. Other factors such as insulin (McNaughton et al., 2008), leptin (Mohammadzadeh and Zarghami,

14 2013; Pasco et al., 2008), ghrelin, level of obesity (Beydoun and Wang, 2010), inflammation and oxidative stress that have been reported for Type 2 diabetes may also play a role as they are known to impact on the risk for depression (Pasco et al., 2010).

Strengths and Limitations The findings of this study should consider the nature of the data utilized and the study limitations. NHANES is a cross-sectional study, restricting the ability to infer the direction of the relationship between diet, depression and diabetes. Even though the data used for the construction of the diet factors were limited to those derived from a 24 hour recall, available research suggests this method can produce accurate and reproducible estimates of the mean intake of groups of individuals if standardized procedures and close interviewer monitoring are used during data gathering (Posner et al., 1982). The self-report instrument of depression, the PHQ-9, may have missed less severe cases of depression (Kroenke and Spitzer, 2002; Kroenke et al., 2001). Even though our measure of diabetes was derived from a number of different sources in the NHANES, without additional clinical data we cannot exclude the possibility that some cases were not identified. The self-reported measure of diabetes had a strong relationship with the medications observed by the interviewer, with diabetic medication usage noted in 68.2% of those stating they had been told they had diabetes. However, there are some participants who reported they had not been told they had diabetes, but in whom diabetic medication usage was documented (0.35%). The severity and type of diabetes may also have influenced its relationship to depression (Mezuk et al., 2013) and the researchers were unable to confirm each case was clinically Type 2 diabetes rather than Type 1 diabetes. However, the average age of diabetes cases in the sample used was approximately 55 years, which is consistent with cases in the general adult population with Type 2 diabetes; it is also the case that Type 2 diabetes accounts for more than 90% of all diabetes cases (Association, 2014). Even though this study acknowledged a possible role of inflammation in the relationship between diabetes and depression by controlling for serum CRP, there is an extensive body of literature to suggest several

15 other possible biological mechanisms for the association (e.g. sympathetic nervous system activation) (Champaneri et al., 2010).

Conclusion The results of this study are concordant with previous research in demonstrating inverse relationships between healthy diet and the probability of depressive symptoms, but also point to the particular importance of healthy diets to those with diabetes. A healthy dietary pattern in this study was associated with a reduced odds of depression, especially among those with Type 2 diabetes. These data suggest that adhering to a healthy diet may be particularly important for avoiding depression among those with diabetes. The results of this study are relevant to the clinical care of people with diabetes and support the current recommendations for the management of Type 2 diabetes with respect to diet. There is evidence to suggest at least a short-term benefit of clinicians teaching those with diabetes strategies to manage their condition (Norris et al., 2001; Norris et al., 2002). The normative role of dietary counselling amongst those with diabetes may be reinforced by highlighting the importance of dietary improvement to the mental health of patients.

16 List of abbreviations BMI – Body Mass Index (kg/m²) CI – Confidence Interval CRP – C-reactive protein DIPIT – Data Integration Protocol In Ten-steps DSM-IV – Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition EFA – Exploratory Factor Analysis KMO – Kaiser-Meyer-Olkin MAP – Minimum Average Partial NHANES – National Health And Nutritional Examination Surveys PCA – Principal Components Analysis PHQ-9 – Patient Health Questionnaire 9 OR – Odds Ratio US – United States WHO – World Health Organisation

17 REFERENCES Akbaraly, T.N., Brunner, E.J., Ferrie, J.E., Marmot, M.G., Kivimaki, M., Singh-Manoux, A., 2009. Dietary pattern and depressive symptoms in middle age. The British Journal of Psychiatry 195, 408-413. Alberti, K.G.M.M., Zimmet, PZ,, 1998. Definition, diagnosis and classification of diabetes mellitus and its complications. Part 1: diagnosis and classification of diabetes mellitus. Provisional report of a WHO consultation. Diabet. Med. 15, 539-553. American Diabetes Association, 2010. Diagnosis and classification of diabetes mellitus. Diabetes Care 33, S62-S69. American Psychiatric Association, 2000. Diagnostic And Statistical Manual Of Mental Disorders DSMIV-TR Fourth Edition (Text Revision) Author: American Psychiatr. Anderson, R.J., Freedland, K.E., Clouse, R.E., Lustman, P.J., 2001. The prevalence of comorbid depression in adults with diabetes a meta-analysis. Diabetes Care 24, 1069-1078. Archer, K.J., Lemeshow, S., 2006. Goodness-of-fit test for a logistic regression model fitted using survey sample data. Stata Journal 6, 97-105. Association, A.D., 2014. Type 1 Diabetes. American Diabetes Association, American Diabetes Association, p. Type 1 Diabetes. Beaton, G.H., Milner, J., Corey, P., McGuire, V., Cousins, M., Stewart, E., De Ramos, M., Hewitt, D., Grambsch, P., Kassim, N., 1979. Sources of variance in 24-hour dietary recall data: implications for nutrition study design and interpretation. Am J Clin Nutr 32, 2546-2559. Berkowitz, S.A., Baggett, T.P., Wexler, D.J., Huskey, K.W., Wee, C.C., 2013. Food insecurity and metabolic control among US Adults With Diabetes. Diabetes Care 36, 3093-3099. Beydoun, M.A., Wang, Y., 2010. Pathways linking socioeconomic status to obesity through depression and lifestyle factors among young US adults. J Affect Disord 123, 52-63. Boyle, J.P., Thompson, T.J., Gregg, E.W., Barker, L.E., Williamson, D.F., 2010. Projection of the year 2050 burden of diabetes in the US adult population: dynamic modeling of incidence, mortality, and prediabetes prevalence. Popul Health Metr 8, 29. Cattell, R., The scientific use of factor analysis in behavioral and life sciences. 1978. Graywind, New York. Cattell, R.B., 1965. A biometrics invited paper. Factor analysis: An introduction to essentials I. The purpose and underlying models. Biometrics 21, 190-215. Cattell, R.B., 1988. The meaning and strategic use of factor analysis, Handbook of multivariate experimental psychology. Springer, pp. 131-203. Centers for Disease Control and Prevention, 2014. National diabetes statistics report: estimates of diabetes and its burden in the United States, 2014. Atlanta, GA: US Department of Health and Human Services, Centers for Disease Control and Prevention. Centers for Disease Control and Prevention National Center for Health Statistics, 2013. National Health and Nutrition Examination Survey: Analytic Guidelines, 1999-2010 U.S. DEPARTMENT OF HEALTH AND HUMAN SERVICES Champaneri, S., Wand, G.S., Malhotra, S.S., Casagrande, S.S., Golden, S.H., 2010. Biological basis of depression in adults with diabetes. Current diabetes reports 10, 396-405. Chocano-Bedoya, P.O., O'Reilly, E.J., Lucas, M., Mirzaei, F., Okereke, O.I., Fung, T.T., Hu, F.B., Ascherio, A., 2013. Prospective study on long-term dietary patterns and incident depression in middle-aged and older women. Am J Clin Nutr 98, 813-820. Committee, I.E., 2009. International Expert Committee report on the role of the A1C assay in the diagnosis of diabetes. Diabetes Care 32, 1327-1334. Danaei, G., Finucane, M., Lu, Y., Singh, G., Cowan, M., Paciorek, C., Lin, J., Farzadfar, F., Khang, Y., Stevens, G., 2011. Global Burden of Metabolic Risk Factors of Chronic Diseases Collaborating Group (Blood Glucose) National, regional, and global trends in fasting plasma glucose and diabetes prevalence since 1980: systematic analysis of health examination surveys and epidemiological studies with 370 country-years and 2.7 million participants. Lancet 378, 31-40. Dipnall, J.F., Berk, M., Jacka, F.N., Williams, L.J., Dodd, S., Pasco, J.A., 2014. Data Integration Protocol In Ten-steps (DIPIT): A new standard for medical researchers. Methods.

18 Dupont, W.D., Dupont, W.D., 2009. Statistical modeling for biomedical researchers: a simple introduction to the analysis of complex data. Cambridge University Press. Eaton, W.W., Hall, A.L., Macdonald, R., Mckibben, J., 2007. Case identification in psychiatric epidemiology: a review. International Review of Psychiatry 19, 497-507. Egede, L.E., Ellis, C., 2010. Diabetes and depression: global perspectives. Diabetes Res. Clin. Pract. 87, 302-312. Egede, L.E., Zheng, D., Simpson, K., 2002. Comorbid depression is associated with increased health care use and expenditures in individuals with diabetes. Diabetes Care 25, 464-470. Field, A., 2009. Discovering statistics using SPSS. Sage publications. Heeringa, S.G., West, B.T., Berglund, P.A., 2010. Applied survey data analysis. CRC Press. Hilbe, J.M., 2009. Logistic regression models. CRC Press Boca Raton. Holt, R.I., de Groot, M., Lucki, I., Hunter, C.M., Sartorius, N., Golden, S.H., 2014. NIDDK international conference report on diabetes and depression: current understanding and future directions. Diabetes Care 37, 2067-2077. Horn, J.L., 1965. A rationale and test for the number of factors in factor analysis. Psychometrika 30, 179185. Hosmer Jr, D.W., Lemeshow, S., Sturdivant, R.X., 2013. Applied logistic regression. John Wiley & Sons. Inzucchi, S.E., 2012. Diagnosis of diabetes. N. Engl. J. Med. 367, 542-550. Jacka, F.N., Cherbuin, N., Anstey, K.J., Butterworth, P., 2014. Dietary Patterns and Depressive Symptoms over Time: Examining the Relationships with Socioeconomic Position, Health Behaviours and Cardiovascular Risk. PLoS One 9, e87657. Jacka, F.N., Kremer, P.J., Berk, M., de Silva-Sanigorski, A.M., Moodie, M., Leslie, E.R., Pasco, J.A., Swinburn, B.A., 2011. A prospective study of diet quality and mental health in adolescents. PLoS One 6, e24805. Jacka, F.N., Kremer, P.J., Leslie, E.R., Berk, M., Patton, G.C., Toumbourou, J.W., Williams, J.W., 2010. Associations between diet quality and depressed mood in adolescents: results from the Australian Healthy Neighbourhoods Study. Australian and New Zealand Journal of Psychiatry 44, 435-442. Kanoski, S.E., Davidson, T.L., 2011. Western diet consumption and cognitive impairment: links to hippocampal dysfunction and obesity. Physiol Behav 103, 59-68. Katon, W.J., Rutter, C., Simon, G., Lin, E.H., Ludman, E., Ciechanowski, P., Kinder, L., Young, B., Von Korff, M., 2005. The association of comorbid depression with mortality in patients with type 2 diabetes. Diabetes Care 28, 2668-2672. Kroenke, K., Spitzer, R.L., 2002. The PHQ-9: a new depression diagnostic and severity measure. Psychiatric Annals 32, 509-515. Kroenke, K., Spitzer, R.L., Williams, J.B., 2001. The PHQǦ9. J Gen Intern Med 16, 606-613. Kroenke, K., Spitzer, R.L., Williams, J.B., Löwe, B., 2010. The patient health questionnaire somatic, anxiety, and depressive symptom scales: a systematic review. Gen Hosp Psychiatry 32, 345-359. Kronish, I.M., Carson, A.P., Davidson, K.W., Muntner, P., Safford, M.M., 2012. Depressive symptoms and cardiovascular health by the american heart association’s definition in the reasons for geographic and racial differences in stroke (REGARDS) study. PLoS One 7, e52771. Lai, J.S., Hiles, S., Bisquera, A., Hure, A.J., McEvoy, M., Attia, J., 2014. A systematic review and metaanalysis of dietary patterns and depression in community-dwelling adults. Am J Clin Nutr 99, 181-197. Lim, S.S., Vos, T., Flaxman, A.D., Danaei, G., Shibuya, K., Adair-Rohani, H., AlMazroa, M.A., Amann, M., Anderson, H.R., Andrews, K.G., 2013. A comparative risk assessment of burden of disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990–2010: a systematic analysis for the Global Burden of Disease Study 2010. The lancet 380, 2224-2260. Lin, E.H., Rutter, C.M., Katon, W., Heckbert, S.R., Ciechanowski, P., Oliver, M.M., Ludman, E.J., Young, B.A., Williams, L.H., McCulloch, D.K., 2010. Depression and Advanced Complications of Diabetes A prospective cohort study. Diabetes Care 33, 264-269. Long, J.S., Freese, J., 2006. Regression models for categorical dependent variables using Stata. Stata press. MacCallum, R.C., Browne, M.W., Sugawara, H.M., 1996. Power analysis and determination of sample size for covariance structure modeling. Psychol. Methods 1, 130. MacCallum, R.C., Widaman, K.F., Zhang, S., Hong, S., 1999. Sample size in factor analysis. Psychol. Methods 4, 84.

19 Martin, A., Rief, W., Klaiberg, A., Braehler, E., 2006. Validity of the brief patient health questionnaire mood scale (PHQ-9) in the general population. Gen Hosp Psychiatry 28, 71-77. Mathers, C.D., Loncar, D., 2006. Projections of global mortality and burden of disease from 2002 to 2030. PLoS Med. 3, e442. McNaughton, S.A., Mishra, G.D., Brunner, E.J., 2008. Dietary patterns, insulin resistance, and incidence of type 2 diabetes in the Whitehall II Study. Diabetes Care 31, 1343-1348. Meyer, B.J., Kolanu, N., Griffiths, D.A., Grounds, B., Howe, P.R., Kreis, I.A., 2013. Food groups and fatty acids associated with self-reported depression: An analysis from the Australian National Nutrition and Health Surveys. Nutrition 29, 1042-1047. Mezuk, B., Eaton, W.W., Albrecht, S., Golden, S.H., 2008. Depression and type 2 diabetes over the lifespan a meta-analysis. Diabetes Care 31, 2383-2390. Mezuk, B., Johnson-Lawrence, V., Lee, H., Rafferty, J.A., Abdou, C.M., Uzogara, E.E., Jackson, J.S., 2013. Is ignorance bliss? Depression, antidepressants, and the diagnosis of prediabetes and type 2 diabetes. Health Psychol. 32, 254. Mohammadzadeh, G., Zarghami, N., 2013. Serum leptin level is reduced in non-obese subjects with type 2 diabetes. International Journal of Endocrinology and Metabolism 11, 3. Murray, C.J., Lopez, A.D., 2013. Measuring the global burden of disease. N. Engl. J. Med. 369, 448-457. Nanri, A., 2013. Nutritional epidemiology of type 2 diabetes and depressive symptoms. J. Epidemiol. 23, 243. Nanri, A., Kimura, Y., Matsushita, Y., Ohta, M., Sato, M., Mishima, N., Sasaki, S., Mizoue, T., 2010. Dietary patterns and depressive symptoms among Japanese men and women. Eur J Clin Nutr 64, 832-839. Nathan, D.M., Kuenen, J., Borg, R., Zheng, H., Schoenfeld, D., Heine, R.J., 2008. Translating the A1C assay into estimated average glucose values. Diabetes Care 31, 1473-1478. Norris, S.L., Engelgau, M.M., Narayan, K.V., 2001. Effectiveness of self-management training in type 2 diabetes a systematic review of randomized controlled trials. Diabetes Care 24, 561-587. Norris, S.L., Lau, J., Smith, S.J., Schmid, C.H., Engelgau, M.M., 2002. Self-Management education for adults with type 2 Diabetes A meta-analysis of the effect on glycemic control. Diabetes Care 25, 11591171. O’connor, B.P., 2000. SPSS and SAS programs for determining the number of components using parallel analysis and Velicer’s MAP test. Behav. Res. Methods Instrum. Comput. 32, 396-402. Pasco, J.A., Jacka, F.N., Williams, L.J., Henry, M.J., Nicholson, G.C., Kotowicz, M.A., Berk, M., 2008. Leptin in depressed women: cross-sectional and longitudinal data from an epidemiologic study. J Affect Disord 107, 221-225. Pasco, J.A., Nicholson, G.C., Williams, L.J., Jacka, F.N., Henry, M.J., Kotowicz, M.A., Schneider, H.G., Leonard, B.E., Berk, M., 2010. Association of high-sensitivity C-reactive protein with de novo major depression. The British Journal of Psychiatry 197, 372-377. Patel, P., Macerollo, A., 2010. Diabetes mellitus: diagnosis and screening. Diabetes 100, 13. Pett, M.A., Lackey, N.R., Sullivan, J.J., 2003. Making sense of factor analysis: The use of factor analysis for instrument development in health care research. Sage. Posner, B.M., Borman, C.L., Morgan, J.L., Borden, W.S., Ohls, J.C., 1982. The validity of a telephoneadministered 24-hour dietary recall methodology. Am J Clin Nutr 36, 546-553. Psaltopoulou, T., Sergentanis, T.N., Panagiotakos, D.B., Sergentanis, I.N., Kosti, R., Scarmeas, N., 2013. Mediterranean diet, stroke, cognitive impairment, and depression: A metaǦanalysis. Ann Neurol 74, 580591. Reynolds, C.F., 3rd, Thomas, S.B., Morse, J.Q., Anderson, S.J., Albert, S., Dew, M.A., Begley, A., Karp, J.F., Gildengers, A., Butters, M.A., Stack, J.A., Kasckow, J., Miller, M.D., Quinn, S.C., 2014. Early intervention to preempt major depression among older black and white adults. Psychiatr Serv 65, 765-773. Rienks, J., Dobson, A., Mishra, G., 2012. Mediterranean dietary pattern and prevalence and incidence of depressive symptoms in mid-aged women: results from a large community-based prospective study. Eur J Clin Nutr 67, 75-82. Roglic, G., Unwin, N., 2010. Mortality attributable to diabetes: estimates for the year 2010. Diabetes Res. Clin. Pract. 87, 15-19. Samieri, C., Jutand, M.-A., Féart, C., Capuron, L., Letenneur, L., Barberger-Gateau, P., 2008. Dietary patterns derived by hybrid clustering method in older people: association with cognition, mood, and selfrated health. J Am Diet Assoc 108, 1461-1471.

20 Sánchez-Villegas, A., Delgado-Rodríguez, M., Alonso, A., Schlatter, J., Lahortiga, F., Majem, L.S., Martínez-González, M.Á., 2009. Association of the Mediterranean dietary pattern with the incidence of depression: the Seguimiento Universidad de Navarra/University of Navarra follow-up (SUN) cohort. Arch Gen Psychiatry 66, 1090-1098. Sánchez-Villegas, A., Martínez-González, M.A., Estruch, R., Salas-Salvadó, J., Corella, D., Covas, M.I., Arós, F., Romaguera, D., Gómez-Gracia, E., Lapetra, J., 2013. Mediterranean dietary pattern and depression: the PREDIMED randomized trial. BMC Med. 11, 208. Seligman, H.K., Bindman, A.B., Vittinghoff, E., Kanaya, A.M., Kushel, M.B., 2007. Food insecurity is associated with diabetes mellitus: results from the National Health Examination and Nutrition Examination Survey (NHANES) 1999–2002. J Gen Intern Med 22, 1018-1023. Spitzer, R.L., Kroenke, K., Williams, J.B., Patient Health Questionnaire Primary Care Study Group, 1999. Validation and utility of a self-report version of PRIME-MD: the PHQ primary care study. JAMA 282, 1737-1744. Strine, T., Mokdad, A., Balluz, L., Gonzalez, O., Crider, R., Berry, J., Kroenke, K., 2008. Depression and anxiety in the United States: findings from the 2006 behavioral risk factor surveillance system. Psychiatr Serv 59, 1383-1390. Tabachnick, B.G., Fidell, L., 2012. Using Multivariate Statistics: International Edition. Pearson. Thompson, F.E., Byers, T., 1994. Dietary assessment resource manual. J Nutr 124, 2245s-2317s. Thurstone, L.L., 1947. Multiple factor analysis. Velicer, W.F., 1976. Determining the number of components from the matrix of partial correlations. Psychometrika 41, 321-327. Velicer, W.F., Eaton, C.A., Fava, J.L., 2000. Construct explication through factor or component analysis: A review and evaluation of alternative procedures for determining the number of factors or components, Problems and solutions in human assessment. Springer, pp. 41-71. Velicer, W.F., Peacock, A.C., Jackson, D.N., 1982. A comparison of component and factor patterns: A Monte Carlo approach. Multivariate Behavioral Research 17, 371-388. Watkins, M.W., 2005. Determining parallel analysis criteria. Journal of Modern Applied Statistical Methods 5, 8. Weber, B., Schweiger, U., Deuschle, M., Heuser, I., 2000. Major depression and impaired glucose tolerance. Experimental and clinical endocrinology & diabetes 108, 187-190. Wilde, P.E., Peterman, J.N., 2006. Individual weight change is associated with household food security status. J Nutr 136, 1395-1400. Wolter, K., 2007. Introduction to variance estimation. Springer. World Health Organisation (WHO), 1999. Definition, diagnosis and classification of diabetes mellitus and its complications. Part.

21 Figures Figure 1: Scree plot for the 25 weighted dietary items Figure 2: Relationship between Diet Factors on Probability of Depression by Gender Figure 3: Relationship between Diet Factors on Probability of Depression by Diabetes Status Tables Table 1: Dietary items Table 2: Weighted Demographic and clinical characteristics by depression¹ status, aged 18 years & above. Table 3: Rotated coefficients for a 5Ǧfactor solution of the Diet Questionnaire using principal components extraction and varimax rotation¹ Table 4: Bivariate Relationships between Depression and Covariates¹. Table 5: Final Multivariate Logistic Regression Model for Depression¹ Table 6: Multivariate Logistic Regression Model stratified by Diabetes Status¹

Highlights x

Diabetes and depression are common comorbid high-prevalence chronic disorders

x

Diet is a risk factor for diabetes and possibly for common mental disorders

x

Five dietary patterns were found using exploratory factor analysis

x

People with diabetes and a healthy diet had lower odds for depressive symptoms

22 Tables Table 1: Dietary items Item: How often eat/drink…..? Cold or hot breakfast Milk or on breakfast Regular soft drinks 100% fruit juice Sweetened coffee/tea Fruit/sports/energy Fruit Leafy/lettuce salad Fried potatoes Non-fried potatoes Beans Other vegetables Pizza

Tomato-based salsa Tomato sauce Red meat Processed meat Cheese Whole grain bread Cooked whole grains Chocolate or candy Pastries Cookies/cake Ice cream Popcorn

23 Table 2: Weighted Demographic and clinical characteristics by depression¹ status, aged 18 years & above.

Total Diabetes¹ No Diabetes Diabetes Gender¹ Male Female Age (yr)¹ 18-34 35-44 45-54 55-64 65+ Marital Status¹ Never Married Married/living with partner Widowed/Divorced/Separated Education¹ Grades 11 and below High School / GED Equivalent Some College/AA / College or Above Race/Ethnicity¹ Mexican American/Hispanic Non-Hispanic White Non-Hispanic Black Other Race - Including Multi-Racial Smoking¹ Current Former Never Adult Food Insecurity¹ Full food security: 0 Marginal food security: 1-2 Low food security: 3-5 Very low food security: 6-10 Body Mass Index (BMI) Category¹ Normal Underweight Overweight Obese Ratio of family income to poverty² C-Reactive Protein (CRP) (mg/dL)²

Not Depressed

Depressed

Total

91.91

8.09

100.00

p-value

92.81 7.19

85.83 14.17

92.26 7.74

0.002

51.64 48.36

36.46 63.54

50.41 49.59

<0.001

33.53 20.66 21.58 17.35 6.87

29.95 21.49 27.26 18.91 2.39

33.24 20.73 22.04 17.48 6.51

0.346 0.032 0.168 0.003

19.25 66.32 14.43

23.46 50.01 26.53

19.60 64.98 15.42

0.006 0.006

15.93 22.59 61.48

30.84 22.39 46.77

17.15 22.57 60.27

<0.001 <0.001

14.40 67.67 11.04 6.89

18.55 59.72 16.58 5.15

14.74 67.03 11.48 6.75

0.004 <0.001 0.517

20.12 22.78 57.10

41.80 16.67 41.54

21.89 22.28 55.83

<0.001 0.968 -

78.60 9.60 6.43 5.38

49.91 14.66 17.26 18.17

76.29 10.01 7.30 6.40

0.001 <0.001 <0.001

29.84 1.79 33.58 34.79 3.11(1.63) 0.35 (0.67)

25.70 2.43 25.73 46.15 2.14(1.84) 0.45 (0.71)

29.51 1.84 32.94 35.71 3.03(1.67) 0.36 (0.67)

0.457 0.470 0.025 <0.001 0.021

¹ Weighted percentages. ² BMI categories: Underweight BMI < 18.5; Normal BMI 18.5 to 24.9, Overweight BMI 25.0 to 29.9; and Obese BMI  30.0. ³ Weighted mean and standard deviation in brackets. Note: Mental health respondents in the 5-17 years age group are not included. P-values from simple binary logistic regression using survey sample estimation, with p-values less than =0.05 in bold. Reference category in italics

24 Table 3: Rotated coefficients for a 5Ǧfactor solution of the Diet Questionnaire using principal components extraction and varimax rotation¹ Factor Name

Healthy Diet

Unhealthy Diet

Sweets Diet

Mexican Diet

Item: How often eat/drink?

Factor 1

Factor 2

Factor 3

Factor 4

Factor 5

h²

Other vegetables

.730

.025

-.012

.027

.054

.624

Leafy/lettuce salad

.665

-.016

-.009

.020

-.017

.624

Fruit

.653

-.090

.063

.052

.193

.228

Cooked whole grains

.427

-.046

-.011

.106

-.125

.478

Whole grain bread

.421

.140

-.020

-.017

.096

.443

Fried potatoes

-.135

.649

.016

.025

.090

.449

Cheese

.256

.507

.059

.072

-.014

.233

Red meat

.011

.492

.105

-.044

-.161

.640

Processed meat

.019

.477

.107

.027

-.079

.537

Pizza

-.151

.474

.121

.067

.139

.287

Non-fried potatoes

.151

.432

.010

.000

.153

.598

Regular soft drinks

-.285

.361

.084

.086

-.046

.281

Cookies/cake

.077

.098

.733

-.005

-.004

.246

Chocolate or candy

-.007

.084

.638

-.040

.011

.332

Ice cream

.035

.093

.576

-.082

.120

.207

Pastries

-.155

.117

.512

.210

.012

.211

Beans

.056

-.044

.013

.797

-.016

.416

Tomato-based salsa

.026

.023

-.015

.769

-.073

.345

Tomato sauce

.102

.224

.009

.415

.160

.553

Milk or on breakfast

.031

.126

.013

.023

.779

.362

Cold or hot breakfast

.109

-.109

.136

.001

.763

.259

Breakfast Diet

¹Significant loadings (>0.3) are in bold red italics; h² represents communalities. Principal components extraction with varimax rotation and Kaiser normalisation. Unweighted sample of 4,643 weighted to the NHANES MEC two year weights.

25 Table 4: Bivariate Relationships between Depression and Covariates¹. Bivariate Logistic

Depressed Healthy Diet Sweets Diet Unhealthy Diet Mexican Diet Breakfast Diet Diabetes Status No Yes Gender Male Female Age Group 18-34 35-44 45-54 55-64 65+ Marital Status Never Married Married/living with partner Widowed/Divorced/Separated Education Grades 11 and below High School / GED Equivalent Some College/AA / College or Above Race/Ethnicity Non-Hispanic White Mexican American/Hispanic Non-Hispanic Black Other Smoking Status Never Current Former Adult Food Insecurity Full food security: 0 Marginal food security: 1-2 Low food security: 3-5 Very low food security: 6-10 Body Mass Index (BMI) Category¹ Normal Underweight Overweight Obese Ratio of family income to poverty C-reactive protein(mg/dL)

95% CI Lower Higher 0.59 0.80 1.00 1.33 0.96 1.27 0.98 1.20 0.92 1.26

Odds Ratio 0.68 1.15 1.10 1.09 1.08

p-value <0.001 0.045 0.161 0.099 0.337

1.00 2.13

0.002

1.38

3.30

1.00 1.86

<0.001

1.40

2.47

1.00 1.16 1.41 1.22 0.39

0.346 0.032 0.168 0.003

0.84 1.03 0.91 0.22

1.62 1.93 1.63 0.70

1.00 0.62 1.51

0.006 0.006

0.45 1.15

0.86 1.98

1.00 0.51 0.39

<0.001 <0.001

0.38 0.27

0.69 0.58

1.00 1.46 1.70 0.85

0.004 <0.001 0.517

1.15 1.32 0.50

1.85 2.20 1.44

1.00 2.86 1.01

<0.001 0.968

2.36 0.75

3.46 1.35

1.00 2.41 4.23 5.32

0.001 <0.001 <0.001

1.49 2.68 3.81

3.89 6.67 7.43

1.00 1.57 0.89 1.54 0.70 1.17

0.457 0.470 0.025 <0.001 0.021

0.45 0.64 1.06 0.63 1.03

5.55 1.24 2.23 0.78 1.34

Reference categories highlighted in italics. Significant covariates bold, p<0.05.

26 Table 5: Final Multivariate Logistic Regression Model for Depression¹ Multivariate Interaction Model

Depressed Diabetes Status No Yes Healthy Diet Interaction Diabetes Status & Healthy Diet No Yes Sweets Diet Gender Male Female Age Group (yr) 18-34 35-44 45-54 55-64 65+ Marital Never Married Married/living with partner Widowed/Divorced/Separated Education Grades 11 and below High School / GED Equivalent Some College/AA / College or Above Race Non-Hispanic White Mexican American/Hispanic Non-Hispanic Black Other Smoking Status Never smoked Current smoker Former smoker Adult Food Insecurity Full food security: 0 Marginal food security: 1-2 Low food security: 3-5 Very low food security: 6-10 Ratio of family income to poverty (PIR) C-reactive protein(mg/dL) Constant

95% CI Lower Higher

Odds Ratio

p-value

1.00 2.29 0.80

0.001 0.050

1.50 0.65

3.51 1.00

1.00 0.81 1.14

0.048 0.237

0.66 0.91

1.00 1.42

1.00 1.89

0.001

1.35

2.64

1.00 1.37 1.65 1.43 0.51

0.120 0.052 0.082 0.129

0.91 1.00 0.95 0.21

2.06 2.73 2.15 1.24

1.00 0.64 0.98

0.013 0.904

0.46 0.65

0.90 1.48

1.00 0.80 0.95

0.276 0.818

0.52 0.59

1.22 1.53

1.00 0.90 1.02

0.431 0.910

0.67 0.75

1.19 1.38

1.00 1.91 1.13

0.001 0.399

1.38 0.84

2.63 1.50

1.00 1.56 2.36 2.84 0.84 1.04 0.06

0.207 0.006 0.001 0.024 0.629 <0.001

0.76 1.32 1.69 0.72 0.88 0.03

3.21 4.22 4.79 0.97 1.22 0.12

¹Reference categories highlighted in italics. Significant covariates bold, p<0.05. Unweighted sample was 3,779, weighted population 150,252,443 (15 Stratum, 31 Primary Sampling Units).

27 Table 6: Multivariate Logistic Regression Model stratified by Diabetes Status¹ Multivariate Model - Stratified by Diabetes Status

Depressed Healthy Diet Sweets Diet Gender Male Female Age Group (yr) 18-34 35-44 45-54 55-64 65+ Marital Status Never Married Married/living with partner Widowed/Divorced/Separated Education Grades 11 and below High School / GED Equivalent Some College/AA / College or Above Race Non-Hispanic White Mexican American/Hispanic Non-Hispanic Black Other Smoking Status Never smoked Current smoker Former smoker AdultFoodInsecurity Fullfoodsecurity:0 ƒ”‰‹ƒŽˆ‘‘†•‡…—”‹–›ǣͳǦʹ ‘™ˆ‘‘†•‡…—”‹–›ǣ͵Ǧͷ ‡”›Ž‘™ˆ‘‘†•‡…—”‹–›ǣ͸ǦͳͲ Ratio of family income to poverty (PIR) C-reactive protein(mg/dL) Constant

Odds Ratio 0.68 1.17 1.00 2.45 1.00 1.66 0.54 0.66 0.25 1.00 0.94 1.52

With Diabetes 95% CI pvalue Lower Higher 0.006 0.52 0.88 0.436 0.77 1.79

0.096

0.487 0.431 0.531 0.087

0.911 0.265

0.84

0.37 0.11 0.16 0.05

0.31 0.70

Odds Ratio 0.79 1.13

Without Diabetes 95% CI pvalue Lower Higher 0.029 0.64 0.97 0.294 0.89 1.45

7.14

1.00 1.81

0.001

1.35

2.42

7.54 2.70 2.65 1.26

1.00 1.31 1.82 1.50 0.51

0.202 0.027 0.164 0.241

0.85 1.08 0.83 0.16

2.03 3.05 2.69 1.65

2.89 3.31

1.00 0.62 0.96

0.011 0.861

0.44 0.58

0.88 1.60

1.00

1.00

0.54

0.045

0.29

0.98

0.87

0.509

0.55

1.36

0.30

0.004

0.14

0.65

1.16

0.540

0.70

1.94

1.00 0.61 1.61 0.55

0.329 0.232 0.412

0.21 0.71 0.12

1.73 3.65 2.47

1.00 0.98 0.98 0.88

0.869 0.871 0.693

0.71 0.71 0.44

1.34 1.34 1.75

1.00 1.19 0.83

0.390 0.707

0.78 0.29

1.81 2.39

1.00 2.07 1.17

<0.001 0.483

1.48 0.74

2.90 1.83

1.00 1.82 0.53 2.15

0.287 0.234 0.071

0.57 0.18 0.93

5.81 1.58 4.95

1.00 1.54 2.90 3.04

0.277 0.001 0.001

0.68 1.63 1.73

3.47 5.17 5.35

0.59 0.90 0.90

0.014 0.580 0.902

0.39 0.60 0.16

0.89 1.35 5.16

0.87 1.07 0.05

0.062 0.454 0.000

0.75 0.89 0.02

1.01 1.28 0.10

Note: Significant covariates bold, p<0.05. Reference category is highlighted in italics

28 ACKNOWLDEGEMENTS There are no acknowledgements to declare.

Competing interests JFD has no conflicts of interest. JAP has received grant/research support from the NHMRC, Perpetual, Amgen (Europe) GmBH, BUPA Foundation and Arthritis Australia and has received speaker fees from Amgen and Sanofi Aventis. DM has no conflicts of interest. MB has received Grant/Research Support from the NIH, Cooperative Research Centre, Simons Autism Foundation, Cancer Council of Victoria, Stanley Medical Research Foundation, MBF, NHMRC, Beyond Blue, Rotary Health, Geelong Medical Research Foundation, Bristol Myers Squibb, Eli Lilly, Glaxo SmithKline, Meat and Livestock Board, Organon, Novartis, Mayne Pharma, Servier and Woolworths, has been a speaker for Astra Zeneca, Bristol Myers Squibb, Eli Lilly, Glaxo SmithKline, Janssen Cilag, Lundbeck, Merck, Pfizer, Sanofi Synthelabo, Servier, Solvay and Wyeth, and served as a consultant to Astra Zeneca, Bioadvantex, Bristol Myers Squibb, Eli Lilly, Glaxo SmithKline, Janssen Cilag, Lundbeck Merck and Servier. Drs Copolov, MB and Bush are co-inventors of two provisional patents regarding the use of NAC and related compounds for psychiatric indications, which, while assigned to the Mental Health Research Institute, could lead to personal remuneration upon a commercialization event. MB is supported by a NHMRC Senior Principal Research Fellowship 1059660. LJW received grants from Eli Lilly, Pfizer, The University of Melbourne, Deakin University and NHMRC. SD has received grants/research support from the Stanley Medical Research Institute, NHMRC, Beyond Blue, ARHRF, Simons Foundation, Geelong Medical Research Foundation, Fondation FondaMental, Eli Lilly, Glaxo SmithKline, Organon, Mayne Pharma and Servier, speaker’s fees from Eli Lilly, advisory board fees from Eli Lilly and Novartis, and conference travel support from Servier. FNJ has no conflicts of interest.

29 Contributors JFD performed the analysis and drafted the manuscript. JAP, DM, MB, LJW, SD and FNJ critically appraised the manuscript. All authors read and approved the final manuscript.

Authors’ Information JFD is a PhD student with the School of Medicine at Deakin University and sessional academic, lecturing in statistics, with the Department of Statistics, Data Science and Epidemiology, School of Health Sciences, Swinburne University of Technology at Swinburne University of Technology. JAP is an epidemiologist who investigates healthy ageing and the role of body composition in the development of chronic disease and the role of inflammation in linking comorbid physical and mental illhealth. She heads the Epi-Centre for Healthy Ageing in the IMPACT Strategic Research Centre at Deakin University. DM is currently Associate Professor at the Department of Statistics, Epidemiology and Data Science, School of Health Sciences, Swinburne University of Technology. MB is currently a NHMRC Senior Principal research Fellow, and is Alfred Deakin Chair of Psychiatry at Deakin University, where he heads the IMPACT Strategic Research Centre. He also is an Honorary Professorial Research fellow in the Department of Psychiatry, the Florey Institute for Neuroscience and Mental Health and Orygen, The National Centre of Excellence in Youth Mental Health, as well as in the School of Public Health and Preventive Medicine at Monash University. SD is a Clinical Associate Professor with the School of Medicine at Deakin University and a Senior Fellow with the Department of Psychiatry at the University of Melbourne. He is part of IMPACT Strategic Research Centre where he is involved with numerous mental health research project. His research interests include; pharmacotherapy, drug safety and the biological underpinnings of mental disorders. FNJ is a psychiatric epidemiologist with a research focus on the prevention of mental disorders

30 LJW is currently a NHMRC Career Development Fellow and Head of the Psychiatric Epidemiology group within the IMPACT Strategic Research Centre, School of Medicine at Deakin University.

ROLE OF FUNDING SOURCE There was no funding source involved with this manuscript.

Figure 1

Figure 2a - to be included with Figure 2b

Figure 2b - to be included with figure 2a

Figure 3a - to be included with figure 3b

Figure 3b - to be included with figure 3a