- Email: [email protected]
Dietary patterns derived by reduced rank regression (RRR) and depressive symptoms in Japanese employees: The Furukawa nutrition and health study
Psychiatry Research ∎ (∎∎∎∎) ∎∎∎–∎∎∎
Contents lists available at ScienceDirect
Psychiatry Research journal homepage: www.elsevier.com/locate/psychres
Dietary patterns derived by reduced rank regression (RRR) and depressive symptoms in Japanese employees: The Furukawa nutrition and health study Takako Miki a,b,n, Takeshi Kochi c, Keisuke Kuwahara a,d, Masafumi Eguchi c, Kayo Kurotani a, Hiroko Tsuruoka c, Rie Ito c, Isamu Kabe c, Norito Kawakami b, Tetsuya Mizoue a, Akiko Nanri a a
Department of Epidemiology and Prevention, Center for Clinical Sciences, National Center for Global Health and Medicine, Tokyo, Japan Department of Mental Health, Graduate School of Medicine, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan Department of Health Administration, Furukawa Electric Corporation, Tokyo, Japan d Teikyo University Graduate School of Public Health, Tokyo, Japan b c
art ic l e i nf o
a b s t r a c t
Article history: Received 27 January 2015 Received in revised form 2 June 2015 Accepted 12 July 2015
Depression has been linked to the overall diet using both exploratory and pre-defined methods. However, neither of these methods incorporates specific knowledge on nutrient-disease associations. The aim of the present study was to empirically identify dietary patterns using reduced rank regression and to examine their relations to depressive symptoms. Participants were 2006 Japanese employees aged 19–69 years. Depressive symptoms were assessed using the Center for Epidemiologic Studies Depression Scale. Diet was assessed using a validated, self-administered diet history questionnaire. Dietary patterns were extracted by reduced rank regression with 6 depression-related nutrients as response variables. Logistic regression was used to estimate odds ratios of depressive symptoms adjusted for potential confounders. A dietary pattern characterized by a high intake of vegetables, mushrooms, seaweeds, soybean products, green tea, potatoes, fruits, and small fish with bones and a low intake of rice was associated with fewer depressive symptoms. The multivariable-adjusted odds ratios of having depressive symptoms were 0.62 (95% confidence interval, 0.48–0.81) in the highest versus lowest tertiles of dietary score. Results suggest that adherence to a diet rich in vegetables, fruits, and typical Japanese foods, including mushrooms, seaweeds, soybean products, and green tea, is associated with a lower probability of having depressive symptoms. & 2015 Published by Elsevier Ireland Ltd.
Keywords: Depression Dietary pattern Epidemiology Japanese Reduced rank regression
1. Introduction Depression, a common condition in the general population, reduces work productivity, lowers quality of life, and increases mortality (Doris et al., 1999). Previous epidemiologic evidence suggested that several kinds of nutrients and foods, such as n-3 polyunsaturated fatty acids, folate, other vitamins, minerals, fruits, and vegetables, are beneficial against depression, but the results have been generally inconclusive (Murakami and Sasaki, 2010). In addition, the investigation of foods or nutrients in isolation may not be a suitable way to assess the interactive or synergic effects n Corresponding author at: Department of Mental Health, Graduate School of Medicine, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan. Fax: þ81 3 5841 3592. E-mail address: [email protected] (T. Miki).
among nutrients (Hu, 2002). In daily life, we do not consume foods or nutrients in isolation but rather in combination as meals. This fact highlights the importance of exploring dietary patterns associated with depression. Dietary pattern analysis, which assesses the effect of overall diet on disease risk, has emerged as a complement to traditional approaches (Hu, 2002). This method may be particularly useful when multiple nutrients are involved in the development of a disease (Hu, 2002). Two types of technique have been commonly used to examine the association between dietary pattern and depression: a priori defined methods (e.g., diet quality score) and exploratory methods (e.g., principal component analysis). In their recent meta-analysis of dietary patterns derived by these methods, Lai et al. suggested that a healthy diet was associated with reduced odds of depression. Besides, the authors also found significant heterogeneity among studies, making it difficult to obtain a
http://dx.doi.org/10.1016/j.psychres.2015.07.033 0165-1781/& 2015 Published by Elsevier Ireland Ltd.
Please cite this article as: Miki, T., et al., Dietary patterns derived by reduced rank regression (RRR) and depressive symptoms in Japanese employees: The Furukawa nutrition and health study. Psychiatry Research (2015), http://dx.doi.org/10.1016/j. psychres.2015.07.033i
T. Miki et al. / Psychiatry Research ∎ (∎∎∎∎) ∎∎∎–∎∎∎
2
reliable summary estimate of the association between dietary patterns and depression (Lai et al., 2014). These methods have advantages and limitations. Diet quality score, which is calculated based on existing knowledge such as recommended diet, captures only limited aspects of the diet and does not consider the correlation structure of food and nutrient intakes (Hoffmann et al., 2004). In contrast, an exploratory statistical approach determines dietary patterns based solely on inter-correlations among dietary variables. However, this method does not incorporate any existing knowledge of nutrient-disease associations and may not capture dietary patterns relevant to disease risk (Hoffmann et al., 2004). Neither method aims to identify dietary patterns that are specific to a disease of interest (Michels and Schulze, 2005). To overcome these issues, a new a posteriori method has been proposed, reduced rank regression (RRR) (Hoffmann et al., 2004). This technique identifies dietary patterns based on several nutrients or biomarkers that have been linked to the disease of interest. This combination of disease-specific knowledge and dietary information of the study population suggests that RRR is an appropriate and promising statistical method to determine dietary patterns associated with an outcome of interest (Hoffmann et al., 2004). To date, RRR has been applied to explore the relationship between an inflammation-prone dietary pattern and depression risk (Lucas et al., 2014). To our knowledge, however, it has not been used to identify dietary patterns based on knowledge of nutrient-mood associations. Here, we used RRR to identify dietary patterns and investigate their association with depressive symptoms among Japanese employees.
2. Methods 2.1. Study procedure and participants Data for the present study was derived from the Furukawa Nutrition and Health Study, which has been described in detail elsewhere (Nanri et al., 2014). A nutritional epidemiological survey was conducted during periodic health checks in April 2012 (factory A) and May 2013 (factory B) among employees of a manufacturing company and their affiliated companies located in Japan's Kanto region. Of 2828 health examination attendants (11% women), 2162 employees agreed to participate in the study (response rate, 76%). We assessed lifestyle, including diet and health status, and obtained health examination data. The protocol of the study was approved by the ethics committee of the National Center for Global Health and Medicine, Japan, and secondary analysis of the Furukawa Nutrition and Health Study data was approved by the ethics committee of the University of Tokyo. Written informed consent was obtained from each participant. We excluded 100 participants with a history of cancer (n ¼20); cardiovascular disease (n¼ 25); chronic hepatitis (n ¼2); kidney disease, including nephritis (n ¼11); pancreatitis (n ¼3); and mental disorders, such as depression and neurotic disorder (n ¼45). Some participants had two or more these conditions. We excluded these participants because such diseases might affect dietary habits or depressive symptoms and thereby cause reverse causality. Of the remaining 2062, we excluded 11 individuals who did not return the study questionnaires and 33 participants who had missing data on the outcome and covariates of the present analysis. We additionally excluded 12 participants with extreme total energy intake ( 43 standard deviations). Finally, 2006 participants (1792 men and 214 women) aged 19–69 years old were included in the analysis.
2.2. Depressive symptoms Depressive symptoms were assessed using a Japanese version (Shima et al., 1985) of the Center for Epidemiologic Studies Depression (CES-D) scale (Radloff, 1977). This scale consists of 20 items addressing 6 typical symptoms of depression experienced during the preceding week, including depressed mood, guilt or worthlessness, helplessness or hopelessness, psychomotor retardation, loss of appetite, and sleep disturbance. Each item is scored on a scale of 0–3 according to the frequency of the symptom, and the scores are then summed to give the total CES-D score, ranging from 0 to 60. The criterion validity of the CES-D scale has been well established in both Western (Radloff, 1977) and Japanese (Shima et al., 1985) subjects. Participants with a CES-D score Z 16 were considered to have depressive symptoms (Radloff, 1977). Another cutoff of Z 19, which might be suitable for Japanese workers (Wada et al., 2007), was also used. 2.3. Dietary assessment Dietary habits during the preceding one-month period were assessed using a validated brief self-administered diet history questionnaire (BDHQ) (Kobayashi et al., 2012) consisting of five sections: (1) intake frequency of 46 food and non-alcoholic beverage items; (2) daily intake of rice and miso soup; (3) frequency of alcoholic drinking and amount of consumption of five alcoholic beverages per typical drinking occasion; (4) usual cooking method; and (5) general dietary behavior. Dietary intakes for 58 food and beverage items, energy, and selected nutrients were estimated using an ad hoc computer algorithm for the BDHQ (Kobayashi et al., 2011), with reference to the Standard Tables of Food Composition in Japan (Science and Technology Agency, 2005a; Science and Technology Agency, 2010b). According to a validation study of the BDHQ using 16-day weighted dietary records as standard, correlation coefficients were 40.40 for the intake of many foods, beverages (Kobayashi et al., 2011), and nutrients used in our study (Kobayashi et al., 2012). 2.4. Other variables The survey questionnaire assessed marital status, job grade, night and rotating shift work, overtime work, smoking, physical activity during work and housework or in commuting to work, and leisure-time physical activity. Physical activity during work and housework or in commuting and leisure-time were expressed as the sum of metabolic equivalents (METs) multiplied by the duration of time (in hours) across physical activities with different levels. 2.5. Statistical analysis Dietary patterns were determined using RRR techniques (Hoffmann et al., 2004). RRR identifies linear functions of food groups (i.e., the dietary patterns) that explain as much of the variation of selected nutrients that are potentially protective or contributing factors to the relevant disease as possible. Therefore, the score for dietary patterns analyzed by this method is likely to be associated with the disease. We selected the following six nutrients as potentially protective factors for depression: folate (Gilbody et al., 2007), vitamin C (Woo et al., 2006; Oishi et al., 2009), magnesium (Jacka et al., 2009, 2012; Yary et al., 2013), calcium (Bae and Kim, 2012), iron (Woo et al., 2006), and zinc (Amani et al., 2010; Jacka et al., 2012; Maserejian et al., 2012; Yary and Aazami, 2012; Vashum et al., 2014). In addition, we found that these nutrients were inversely associated with depressive symptoms in a preliminary analysis in our study population. Dietary
Please cite this article as: Miki, T., et al., Dietary patterns derived by reduced rank regression (RRR) and depressive symptoms in Japanese employees: The Furukawa nutrition and health study. Psychiatry Research (2015), http://dx.doi.org/10.1016/j. psychres.2015.07.033i
T. Miki et al. / Psychiatry Research ∎ (∎∎∎∎) ∎∎∎–∎∎∎
3
Table 1 Explained percentage of variations in nutrients (response variables) and food and beverage items by extracted dietary patterns.a
Explained variation in nutrients (%) Explained variation in food and beverage items (%) Pearson correlation coefficients b Folate Vitamin C Magnesium Calcium Iron Zinc
DP1
DP2
DP3
DP4
DP5
DP6
Total explained variation
73.2 8.6
11.5 2.4
7.3 2.1
4.4 2.8
2.3 1.8
1.1 2.2
99.9 19.9
0.92c 0.84c 0.90c 0.80c 0.94c 0.71c
0.29c 0.40c 0.02 0.21c 0.003 0.63c
0.12c 0.10c 0.17c 0.52c 0.19c 0.29c
0.06 0.32c 0.33c 0.16c 0.13c 0.11c
0.19c 0.16c 0.21c 0.10c 0.13c 0.06
0.16c 0.03 0.03 0.01 0.18c 0.06
Abbreviation: DP, Dietary Pattern. a b c
Dietary patterns were derived by reduced rank regression analysis. Pearson correlation coefficients between nutrients (response variables) and extracted dietary patterns. Po 0.001.
patterns related to the intake of these 6 nutrients were derived on the basis of 52 food and beverage items, excluding 6 items (sugar added to coffee and black tea, three items usually added during cooking [salt, oil, and sugar], table salt and salt-containing seasoning at the table, and soup consumed with noodles). The 6 nutrients selected as response variables and 52 food and beverage items were energy-adjusted prior to the RRR using the density method. Although n-3 polyunsaturated fatty acid levels have been shown to be inversely associated with depressive symptoms (Murakami and Sasaki, 2010), the present and previous studies among Japanese (Miyake et al., 2006; Murakami et al., 2008) found no significant association between them. Accordingly, we decided not to include n-3 polyunsaturated fatty acids as response variables. Explained percentage of variations in nutrients (response variables) and food and beverage items by extracted dietary patterns and Pearson correlation coefficients between nutrients and dietary patterns were calculated. Of the six dietary patterns extracted, we focused on dietary pattern 1 (DP1) in our statistical analysis because it explains the largest amount of variation among the nutrients. Moreover, when dietary patterns were extracted on the basis of a reduced set of food groups (22 food groups), similar dietary patterns were observed. We calculated factor loadings, which represent the magnitude and direction of each food and beverage item's contribution to DP1 scores, of food and beverage items associated with DP1 and Pearson correlation coefficients between food and beverage items and nutrients (response variables). A positive factor loading value indicated that a higher DP1 score was associated with an increased intake for that food or beverage items. On the other hand, a negative factor loading value indicated less intake of the items. To show the general characteristics of the study population, the mean or proportion was presented according to the tertile of DP1 score. Trend association across tertile categories of scores for DP1 was assessed using the Mantel–Haenszel χ2-test for categorical variables and linear regression analysis for continuous variables, assigning ordinal numbers 1–3 to tertile categories of DP1 scores. To examine the association between DP1 and depressive symptoms, we performed a multiple logistic regression analysis and calculated the odds ratios (ORs) and 95% confidence intervals (CIs) of depressive symptoms for the tertiles of DP1 scores, using the lowest tertile category as reference. The first model was adjusted for age (in years, continuous), sex, and site (survey in April 2012 or in May 2013). The second model was further adjusted for marital status (married or other), job grade (low, middle, or high), night or rotating shift work (yes or no), overtime work (o10 h/month, 10– o30 h/month, or Z30 h/month), physical activity at work and housework or in commuting to work (o3 METs-h/day, 3– o7 METs-h/day, 7– o20 METs-h/day, or Z20 METs-h/day),
leisure-time physical activity (not engaged, 4 0– o3 METs-h/ week, 3– o10 METs-h/week, or Z10 METs-h/week), smoking (never-smoker, quitter, current smoker consuming o20 cigarettes/day, or current smoker consuming Z20 cigarettes/day), and total energy intake (kcal/day, continuous). We repeated the analysis after excluding participants with obesity, a history of diabetes, and/or hypertension, on the basis that these conditions might affect dietary habits and may lead to reverse causality. Additionally, we randomly divided the sample in half and analyzed the data in two subsamples to confirm that the results were reproducible in subsamples. Two-sided P o0.05 was considered statistically significant in all analyses. The RRR analyses were performed using Statistical Analysis System (SAS) software version 9.4 (SAS Institute, Cary, NC, USA), while all the other analyses were performed using Stata version 12.1 (StataCorp, College Station, TX, USA).
3. Results DP1 scores were highly correlated with each nutrient (folate, vitamin C, magnesium, calcium, iron, and zinc). All Pearson correlation coefficients were greater than 0.70 (Table 1). Factor loadings of the 52 food and beverage items associated with DP1 scores and correlation coefficients between food and beverage items and the 6 response variables are presented in Table 2. Food items with a factor loading greater than 0.15 were a high intake of vegetables, mushrooms, seaweeds, soybean products, green tea, potatoes, fruits, and small fish with bones. In contrast, the only food item with a factor loading less than –0.15 was rice. Table 3 shows the characteristics of study participants according to tertile of DP1 score. Participants with higher DP1 scores were older, more likely to be women and physically active in leisure, but less likely to be night or rotating shift workers, overtime workers, smokers, and alcohol drinkers. Mean values of folate, vitamin C, magnesium, calcium, iron, and zinc increased with higher DP1 score. The prevalence of depressive symptoms was 27.8%. The ORs of depressive symptoms according to tertile categories of DP1 are shown in Table 4. In an age-, sex-, and workplace-adjusted model (model 1), DP1 was significantly associated with a decreased prevalence of depressive symptoms. Further adjustment for other covariates (model 2) somewhat attenuated the association, but it nevertheless remained significant. The multivariate-adjusted ORs for the lowest through highest tertile of DP1 were 1.00 (reference), 0.78 (95% CI, 0.61–0.99), and 0.62 (95% CI, 0.48–0.81), respectively (P for trend o0.001). When a higher cutoff value for the definition of depressive symptoms was used (CES-D Z19), the ORs for depressive symptoms were 1.00 (reference), 0.93 (95% CI, 0.70–1.23),
Please cite this article as: Miki, T., et al., Dietary patterns derived by reduced rank regression (RRR) and depressive symptoms in Japanese employees: The Furukawa nutrition and health study. Psychiatry Research (2015), http://dx.doi.org/10.1016/j. psychres.2015.07.033i
T. Miki et al. / Psychiatry Research ∎ (∎∎∎∎) ∎∎∎–∎∎∎
4
Table 2 Factor loadings of food and beverage items associated with dietary pattern 1 and correlation coefficients between food and beverage items and nutrients (response variables)a. Food and beverage items
Low-fat milk and yogurt Milk and yogurt Squid/octopus/shrimp/shellfish Small fish with bones Dried fish/salted fish Oily fish Lean fish Egg Tofu/atsuage b Natto c Potatoes Pickled green leaves vegetables Lettuces/cabbage (raw) Green leaves vegetables Cabbage/Chinese cabbage Carrots/pumpkin Japanese radish/turnip Other root vegetables Tomatoes Mushrooms Seaweeds Citrus fruit Persimmons/strawberries/kiwifruit Other fruit Green tea Cola drink/soft drink Rice
Factor loadings
Correlation between food items and 6 response variables
Dietary pattern 1
Folate
Vitamin C
Magnesium
Calcium
Iron
Zinc
0.11 0.11 0.12 0.15 0.14 0.12 0.14 0.13 0.21 0.17 0.16 0.13 0.23 0.33 0.27 0.25 0.21 0.24 0.19 0.24 0.20 0.15 0.14 0.15 0.17 0.11 0.15
0.11 0.12 0.19 0.22 0.19 0.18 0.23 0.23 0.34 0.29 0.30 0.27 0.53 0.76 0.62 0.54 0.43 0.52 0.41 0.48 0.37 0.28 0.25 0.26 0.45 0.21 0.30
0.11 0.16 0.14 0.16 0.15 0.15 0.22 0.14 0.25 0.15 0.38 0.22 0.47 0.65 0.57 0.53 0.40 0.50 0.43 0.44 0.33 0.46 0.47 0.41 0.43 0.16 0.28
0.21 0.15 0.29 0.38 0.34 0.28 0.34 0.18 0.50 0.46 0.34 0.27 0.42 0.61 0.48 0.49 0.42 0.47 0.35 0.47 0.42 0.26 0.22 0.28 0.22 0.24 0.35
0.53 0.49 0.15 0.39 0.30 0.22 0.23 0.20 0.39 0.23 0.20 0.23 0.38 0.53 0.41 0.41 0.33 0.35 0.32 0.37 0.33 0.27 0.26 0.32 0.16 0.16 0.34
0.12 0.07 0.25 0.27 0.36 0.29 0.30 0.36 0.47 0.41 0.32 0.29 0.44 0.67 0.49 0.46 0.39 0.45 0.35 0.46 0.40 0.24 0.22 0.23 0.39 0.22 0.31
0.21 0.21 0.34 0.29 0.23 0.23 0.25 0.32 0.36 0.29 0.23 0.16 0.28 0.37 0.29 0.31 0.25 0.30 0.25 0.33 0.27 0.16 0.15 0.15 0.14 0.22 0.02
a Omitted in the tables were food items with factor loadings less than 70.10 for dietary pattern 1 (chicken, pork/beef, ham/sausage/bacon, liver, canned tuna, other pickles, western-type confectioneries, Japanese confectioneries, rice crackers/rice cake/okonomiyaki: meat/fish and vegetables pancake, ice cream, mayonnaise/dressing, bread, buckwheat noodles, Japanese wheat noodles, Chinese noodles, spaghetti and macaroni, black tea/oolong tea, coffee, 100% fruit and vegetable juice, miso soup, sake, beer, shochu, whisky, and wine). b Deep-fried tofu. c Fermented soybeans.
and 0.69 (95% CI, 0.51–0.94), respectively (P for trend¼0.019). A stronger association was observed after excluding participants
with obesity, a history of diabetes, and/or hypertension (n ¼1407); the adjusted ORs of depressive symptoms for the lowest through
Table 3 Characteristics of the study population by tertiles of dietary pattern 1 scoresa. Variables
Number of subjects Age (mean7 s.d., year) Sex (women, %) Site (survey in April 2012, %) Marital status (married, %) Job grade (low, %) Night or rotating shift work (yes, %) Overtime work ( Z 30 h/month, %) Physical activity at work and housework or in commuting to work (Z 20 METs-h/day, %) Leisure-time physical activities ( Z10 METs-h/week, %) Smoking status (current, %) Alcohol drinking (currentb, %) Daily dietary intake (mean7 s.d) Total energy (kcal/day) Folate (μg /1000 kcal) Vitamin C (mg/1000 kcal) Magnesium (mg/1000 kcal) Calcium (mg/1000 kcal) Iron (mg/1000 kcal) Zinc (mg/1000 kcal)
Dietary pattern 1 scores Tertile 1 (low)
Tertile 2
Tertile 3 (high)
P-trenda
669 40.5 7 9.3 5.8 52.0 62.6 71.5 31.5 28.4 30.8 20.8 39.8 57.4
669 42.17 9.9 9.9 55.5 68.8 68.3 17.9 27.2 20.5 27.7 25.6 55.9
668 44.0 7 10.7 16.3 59.9 66.6 67.4 10.9 24.1 17.7 29.9 21.1 45.7
o 0.001 o 0.001 0.015 0.056 0.14 o 0.001 o 0.001 o 0.001 o 0.001 o 0.001 o 0.001
1785 7 511 1137 26 317 12 102 7 14 1657 49 2.9 7 0.5 3.7 7 0.5
1836 7 467 1567 25 48 7 12 1237 12 2277 52 3.7 70.4 4.2 70.4
1768 7488 222 7 60 727 25 1527 23 3107 84 4.8 7 0.9 4.6 7 0.5
0.69 o 0.001 o 0.001 o 0.001 o 0.001 o 0.001 o 0.001
Abbreviations: s.d., standard deviation; METs; Metabolic Equivalents a On the basis of the Mantel–Haenszel χ2-test for categorical variables and linear regression analysis for continuous variables, assigning ordinal numbers 1–3 to tertile categories of scores for dietary pattern 1. b Alcohol consumption of at least one day per week.
Please cite this article as: Miki, T., et al., Dietary patterns derived by reduced rank regression (RRR) and depressive symptoms in Japanese employees: The Furukawa nutrition and health study. Psychiatry Research (2015), http://dx.doi.org/10.1016/j. psychres.2015.07.033i
T. Miki et al. / Psychiatry Research ∎ (∎∎∎∎) ∎∎∎–∎∎∎ Table 4 Odds ratios and 95% confidence intervals for depressive symptoms according to tertiles of dietary pattern 1 scores. Variables
CES-D (15/16) Subjects with/without depressive symptoms Model 1b Model 2c CES-D (18/19) Subjects with/without depressive symptoms Model 1b Model 2c
Dietary pattern 1 scores Tertile 1 (low)
Tertile 2
Tertile 3 (high)
234/435
180/489
143/525
1.00 (ref)
0.70 (0.56– 0.89) 0.78 (0.61– 0.99)
0.54 (0.42– 0.69) 0.62 (0.48– 0.81)
150/519
124/545
89/579
1.00 (ref)
0.82 (0.63– 1.07) 0.93 (0.70– 1.23)
0.58 (0.44– 0.78) 0.69 (0.50– 0.94)
1.00 (ref)
1.00 (ref)
P-trenda
o 0.001 o 0.001
o 0.001 0.019
Abbreviations: CES-D, Center for Epidemiologic Studies Depression Scale; ref, reference. a Based on multiple logistic regression analyses, assigning ordinal numbers of 1–3 to the tertile categories of scores for dietary pattern 1. b Adjusted for age (year, continuous), sex, and site (survey in April 2012 or in May 2013). c Adjusted for age (year, continuous), sex, site (survey in April 2012 or in May 2013), marital status (married or other), job grade (low, middle, or high), night or rotating shift work (yes or no), overtime work ( o 10 h/month, 10– o30 h/month, or Z 30 h/month), physical activity at work and housework or in commuting to work ( o 3 METs-h/day, 3– o7 METs-h/day, 7–o 20 METs-h/day, or Z 20 METs-h/ day), leisure-time physical activity (not engaged, 4 0–o 3 METs-h/week, 3–o 10 METs-h/week, or Z 10 METs-h/week), smoking (never-smoker, quitter, current smoker consuming o20 cigarettes/day, or current smoker consuming Z20 cigarettes/day), and total energy intake (kcal/day, continuous).
highest tertile of DP1 in model 2 was 1.00 (reference), 0.63 (95% CI, 0.46–0.84), and 0.50 (95% CI, 0.37–0.69), respectively (P for trend o0.001). In addition, inverse associations were observed in both randomly divided subsamples.
4. Discussion In this study in Japanese workers, RRR identified a dietary pattern that was associated with a lower prevalence of depressive symptoms. This dietary pattern was characterized by a high intake of vegetables, mushrooms, seaweeds, soybean products, green tea, potatoes, fruits, and small fish with bones, as well as a low intake of rice. To our knowledge, this is the first study to investigate the association between dietary pattern based on existing knowledge of nutrients and depressive symptoms using an empirical dietary pattern method, RRR. Dietary patterns similar to the present one (DP1) have been linked to depressive symptoms. In a recent meta-analysis of observational studies that used a priori methods (e.g., diet quality scores or indexes) and a posteriori approaches (e.g., factor analysis, including principal component analysis) in the identification of dietary patterns, a healthy diet pattern, characterized by high intake of fruits, vegetables, fish, and whole grains, which is similar to our present derived dietary pattern, was inversely associated with depression (Lai et al., 2014). The present findings are also consistent with those of two other Japanese studies that identified dietary patterns using principal component analysis: a healthy Japanese dietary pattern characterized by a high intake of vegetables, fruit, mushrooms, and soy products was associated with a decreased prevalence of depressive symptoms (Nanri et al., 2010); and a balanced Japanese dietary pattern featuring a high intake of
5
vegetables, mushrooms, and seaweed was related to a decreased prevalence of depressive symptoms (Suzuki et al., 2013). Even if a dietary pattern extracted using principal component analysis is found to be associated with a specific disease or condition, the interpretation of such findings is often difficult due to their exploratory nature (Hoffmann et al., 2004). In contrast, the association with a dietary pattern derived using an empirical RRR method may provide some insight into mediators linking food intake and disease. Although the dietary pattern in the present study was similar to those identified using principal component analysis (Nanri et al., 2010; Suzuki et al., 2013), our results provide additional evidence that a diet rich in potentially important nutrients (folate, vitamin C, magnesium, calcium, iron, and zinc) may be inversely associated with depressive symptoms. The inverse association between DP1 and depressive symptoms in the present study is well supported by the results of epidemiological studies on the association of depression with folate (Gilbody et al., 2007), vitamin C (Woo et al., 2006; Oishi et al., 2009), magnesium (Jacka et al., 2009, 2012; Yary et al., 2013), calcium (Bae and Kim, 2012), iron (Woo et al., 2006), and zinc (Amani et al., 2010; Jacka et al., 2012; Maserejian et al., 2012; Yary and Aazami, 2012; Vashum et al., 2014). These nutrients may independently as well as jointly protect against depression through multiple pathways. Folate is involved in the metabolism of monoamines in the brain, such as serotonin, and may protect brain function via its ability to reduce homocysteine, which has a neurotoxic effect (Bottiglieri, 2005). Vitamin C (ascorbic acid) is required to recycle tetrahydrobiopterin, which is necessary for tryptophan hydroxylase activity in serotonin synthesis (Meredith and May, 2013). In addition, reactive oxygen species (ROS) and defective antioxidant defenses are thought to be involved in the pathophysiology of depression (Bilici et al., 2001). Thus, antioxidant vitamins, such as vitamin C, may prevent depression. Magnesium is necessary for tryptophan hydroxylase and serotonin receptor binding (Kantak, 1988) and is a potent antagonist of the N-methyl-D-aspartate (NMDA) receptor complex (Bresink et al., 1995). Calcium activates tryptophan hydroxylase in the synthesis of serotonin (Knapp et al., 1975). Iron is involved in the synthesis of neurotransmitters, including dopamine and serotonin (Beard et al., 1993). Zinc might modulate mood by enhancing the function of the serotonergic system or inhibiting the function of the NMDA receptor complex in the central nervous system (Szewczyk et al., 2008). The major strengths of this study include its high participation rate and use of validated methodologies to assess nutrient intake (i.e., the BDHQ). Some limitations should also be noted. First, an association derived from a cross-sectional study design does not necessarily indicate causality. For instance, the lower intake of food among persons with depressed mood than those without may be caused by decreased appetite due to the depressive state. However, this is less likely in our study because total energy intake between participants with and without depressed mood was similar (1827 kcal and 1785 kcal, respectively). Nevertheless, we could not exclude the possibility that the choice of foods determining the dietary pattern had changed as a result of the depressive state. Second, we chose depression-related nutrients as response variables, rather than biomarkers, in RRR. Given that the dietary intake of a nutrient does not necessary reflect body nutritional status, which is also determined by absorption, metabolism, and excretion, an RRR study that includes nutritional biomarkers as response variables is warranted. Third, we assessed depressive symptoms using a validated questionnaire (i.e., the CES-D scale), but not structured diagnostic interviews. The association may differ if the outcome is clinically diagnosed depression. Fourth, although we adjusted for a range of potential confounding variables, we cannot rule out the possibility that the
Please cite this article as: Miki, T., et al., Dietary patterns derived by reduced rank regression (RRR) and depressive symptoms in Japanese employees: The Furukawa nutrition and health study. Psychiatry Research (2015), http://dx.doi.org/10.1016/j. psychres.2015.07.033i
T. Miki et al. / Psychiatry Research ∎ (∎∎∎∎) ∎∎∎–∎∎∎
6
observed associations were due to unmeasured confounders and residual confounding. Finally, because the present study was conducted among workers in a Japanese manufacturing company, caution is required in generalizing the findings. In conclusion, this study suggests that adherence to a diet rich in vegetables, fruits, and typical Japanese foods, including mushrooms, seaweeds, soybean products, and green tea, may be associated with a lower prevalence of depressive symptoms. The observed association in this cross-sectional study requires confirmation in prospective studies among populations with different backgrounds.
Contributors T. Mizoue and A.N. designed the research; T.K., M.E., K. Kuwahara, H.T., K. Kurotani, R. I., I. Kabe, T. Mizoue, and A.N. conducted the research; and T. Miki performed the statistical analysis, wrote the manuscript, and had primary responsibility for the final content. All authors were involved in revision of the manuscript and approved the final version of the manuscript. None of the authors had any conflicts of interest. T.K., M.E., H.T., R.I., and I. Kabe are health professionals in Furukawa Electric Corporation.
Conflict of interest The authors declare no conflict of interest. T.K., M.E., H.T., R.I., and I. Kabe are health professionals in Furukawa Electric Corporation.
Acknowledgments We thank Fumiko Zaizen (Furukawa Electric Corporation) and Ayami Kume, Sachiko Nishihara, Yuho Mizoue, Saeko Takagiwa, and Yuriko Yagi (National Center for Global Health and Medicine) for their help in data collection. This study was supported by a Grant-in-Aid for Young Scientists (A) (25702006) from the Japan Society for the Promotion of Science, Health and Labour Sciences Research Grant, and Health and Labour Sciences Research Expenses for Commission (Comprehensive Research on Life-Style Related Diseases including Cardiovascular Diseases and Diabetes Mellitus H26-005).
References Amani, R., Saeidi, S., Nazari, Z., Nematpour, S., 2010. Correlation between dietary zinc intakes and its serum levels with depression scales in young female students. Biol. Trace Elem. Res. 137, 150–158. Bae, Y.J., Kim, S.K., 2012. Low dietary calcium is associated with self-rated depression in middle-aged Korean women. Nutr. Res. Pract. 6, 527–533. Beard, J.L., Connor, J.R., Jones, B.C., 1993. Iron in the brain. Nutrition Reviews 51, 157–170. Bilici, M., Efe, H., Koroglu, M.A., Uydu, H.A., Bekaroglu, M., Deger, O., 2001. Antioxidative enzyme activities and lipid peroxidation in major depression: alterations by antidepressant treatments. J. Affect. Disord. 64, 43–51. Bottiglieri, T., 2005. Homocysteine and folate metabolism in depression. Prog. Neuro-Psychopharmacol. Biol. Psychiatry 29, 1103–1112. Bresink, I., Danysz, W., Parsons, C.G., Mutschler, E., 1995. Different binding affinities of NMDA receptor channel blockers in various brain regions—indication of NMDA receptor heterogeneity. Neuropharmacology 34, 533–540. Doris, A., Ebmeier, K., Shajahan, P., 1999. Depressive illness. Lancet 354, 1369–1375. Gilbody, S., Lightfoot, T., Sheldon, T., 2007. Is low folate a risk factor for depression? A meta-analysis and exploration of heterogeneity. J. Epidemiol. Commun. Health 61, 631–637. Hoffmann, K., Schulze, M.B., Schienkiewitz, A., Nothlings, U., Boeing, H., 2004. Application of a new statistical method to derive dietary patterns in nutritional epidemiology. Am. J. Epidemiol. 159, 935–944. Hu, F.B., 2002. Dietary pattern analysis: a new direction in nutritional
epidemiology. Curr. Opin. Lipidol. 13, 3–9. Jacka, F.N., Overland, S., Stewart, R., Tell, G.S., Bjelland, I., Mykletun, A., 2009. Association between magnesium intake and depression and anxiety in community-dwelling adults: the Hordaland Health Study. Aust. N. Z. J. Psychiatry 43, 45–52. Jacka, F.N., Maes, M., Pasco, J.A., Williams, L.J., Berk, M., 2012. Nutrient intakes and the common mental disorders in women. J. Affect. Disord. 141, 79–85. Kantak, K.M., 1988. Magnesium deficiency alters aggressive behavior and catecholamine function. Behav. Neurosci. 102, 304–311. Knapp, S., Mandell, A.J., Bullard, W.P., 1975. Calcium activation of brain tryptophan hydroxylase. Life Sci. 16, 1583–1593. Kobayashi, S., Murakami, K., Sasaki, S., Okubo, H., Hirota, N., Notsu, A., Fukui, M., Date, C., 2011. Comparison of relative validity of food group intakes estimated by comprehensive and brief-type self-administered diet history questionnaires against 16 d dietary records in Japanese adults. Public Health Nutr. 14, 1200–1211. Kobayashi, S., Honda, S., Murakami, K., Sasaki, S., Okubo, H., Hirota, N., Notsu, A., Fukui, M., Date, C., 2012. Both comprehensive and brief self-administered diet history questionnaires satisfactorily rank nutrient intakes in Japanese adults. J. Epidemiol. 22, 151–159. Lai, J.S., Hiles, S., Bisquera, A., Hure, A.J., McEvoy, M., Attia, J., 2014. A systematic review and meta-analysis of dietary patterns and depression in communitydwelling adults. Am. J. Clin. Nutr. 99, 181–197. Lucas, M., Chocano-Bedoya, P., Shulze, M.B., Mirzaei, F., O’Reilly, E.J., Okereke, O.I., Hu, F.B., Willett, W.C., Ascherio, A., 2014. Inflammatory dietary pattern and risk of depression among women. Brain Behav. Immun. 36, 46–53. Maserejian, N.N., Hall, S.A., McKinlay, J.B., 2012. Low dietary or supplemental zinc is associated with depression symptoms among women, but not men, in a population-based epidemiological survey. J. Affect. Disord. 136, 781–788. Meredith, M.E., May, J.M., 2013. Regulation of embryonic neurotransmitter and tyrosine hydroxylase protein levels by ascorbic acid. Brain Res. 1539, 7–14. Michels, K.B., Schulze, M.B., 2005. Can dietary patterns help us detect diet-disease associations? Nutr. Res. Rev. 18, 241–248. Miyake, Y., Sasaki, S., Yokoyama, T., Tanaka, K., Ohya, Y., Fukushima, W., Saito, K., Ohfuji, S., Kiyohara, C., Hirota, Y., 2006. Risk of postpartum depression in relation to dietary fish and fat intake in Japan: the Osaka Maternal and Child Health Study. Psychol. Med. 36, 1727–1735. Murakami, K., Mizoue, T., Sasaki, S., Ohta, M., Sato, M., Matsushita, Y., Mishima, N., 2008. Dietary intake of folate, other B vitamins, and omega-3 polyunsaturated fatty acids in relation to depressive symptoms in Japanese adults. Nutrition 24, 140–147. Murakami, K., Sasaki, S., 2010. Dietary intake and depressive symptoms: a systematic review of observational studies. Mol. Nutr. Food Res. 54, 471–488. 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. Nanri, A., Eguchi, M., Kuwahara, K., Kochi, T., Kurotani, K., Ito, R., Pham, N.M., Tsuruoka, H., Akter, S., Jacka, F., Mizoue, T., Kabe, I., 2014. Macronutrient intake and depressive symptoms among Japanese male workers: the Furukawa Nutrition and Health Study. Psychiatry Res. 220, 263–268. Oishi, J., Doi, H., Kawakami, N., 2009. Nutrition and depressive symptoms in community-dwelling elderly persons in Japan. Acta Med. Okayama 63, 9–17. Radloff, L.S., 1977. The CES-D scale: a self-report depression scale for research in the general population. Appl. Psychol. Meas. 1, 385–401. Science and Technology Agency, 2005a. Standard Tables of Food Composition in Japan, fatty acids section. 5th revised and enlarged ed. Printing Bureau of the Ministry of Finance, Tokyo. (in Japanese). Science and Technology Agency, 2010b. Standard Tables of Food Composition in Japan, 2010. Tokyo: Ministry of Education, Culture, Sports, Science and Technology. (in Japanese). Shima, S., Shikano, T., Kitamura, T., Asai, M., 1985. New self-rating scale for depression. Japanese. J. Clin. Psychiatry 27, 717–723. Suzuki, T., Miyaki, K., Tsutsumi, A., Hashimoto, H., Kawakami, N., Takahashi, M., Shimazu, A., Inoue, A., Kurioka, S., Kakehashi, M., Sasaki, Y., Shimbo, T., 2013. Japanese dietary pattern consistently relates to low depressive symptoms and it is modified by job strain and worksite supports. J. Affect. Disord. 150, 490–498. Szewczyk, B., Poleszak, E., Sowa-Kucma, M., Siwek, M., Dudek, D., Ryszewska-Pokrasniewicz, B., Radziwon-Zaleska, M., Opoka, W., Czekaj, J., Pilc, A., Nowak, G., 2008. Antidepressant activity of zinc and magnesium in view of the current hypotheses of antidepressant action. Pharmacol. Rep. 60, 588–589. Vashum, K.P., McEvoy, M., Milton, A.H., McElduff, P., Hure, A., Byles, J., Attia, J., 2014. Dietary zinc is associated with a lower incidence of depression: findings from two Australian cohorts. J. Affect. Disord. 166, 249–257. Wada, K., Tanaka, K., Theriault, G., Satoh, T., Mimura, M., Miyaoka, H., Aizawa, Y., 2007. Validity of the Center for Epidemiologic Studies Depression Scale as a screening instrument of major depressive disorder among Japanese workers. Am. J. Ind. Med. 50, 8–12. Woo, J., Lynn, H., Lau, W.Y., Leung, J., Lau, E., Wong, S.Y., Kwok, T., 2006. Nutrient intake and psychological health in an elderly Chinese population. Int. J. Geriatr. Psychiatry 21, 1036–1043. Yary, T., Aazami, S., 2012. Dietary intake of zinc was inversely associated with depression. Biol. Trace Elem. Res. 145, 286–290. Yary, T., Aazami, S., Soleimannejad, K., 2013. Dietary intake of magnesium may modulate depression. Biol. Trace Elem. Res. 151, 324–329.
Please cite this article as: Miki, T., et al., Dietary patterns derived by reduced rank regression (RRR) and depressive symptoms in Japanese employees: The Furukawa nutrition and health study. Psychiatry Research (2015), http://dx.doi.org/10.1016/j. psychres.2015.07.033i
Contents lists available at ScienceDirect
Psychiatry Research journal homepage: www.elsevier.com/locate/psychres
Dietary patterns derived by reduced rank regression (RRR) and depressive symptoms in Japanese employees: The Furukawa nutrition and health study Takako Miki a,b,n, Takeshi Kochi c, Keisuke Kuwahara a,d, Masafumi Eguchi c, Kayo Kurotani a, Hiroko Tsuruoka c, Rie Ito c, Isamu Kabe c, Norito Kawakami b, Tetsuya Mizoue a, Akiko Nanri a a
Department of Epidemiology and Prevention, Center for Clinical Sciences, National Center for Global Health and Medicine, Tokyo, Japan Department of Mental Health, Graduate School of Medicine, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan Department of Health Administration, Furukawa Electric Corporation, Tokyo, Japan d Teikyo University Graduate School of Public Health, Tokyo, Japan b c
art ic l e i nf o
a b s t r a c t
Article history: Received 27 January 2015 Received in revised form 2 June 2015 Accepted 12 July 2015
Depression has been linked to the overall diet using both exploratory and pre-defined methods. However, neither of these methods incorporates specific knowledge on nutrient-disease associations. The aim of the present study was to empirically identify dietary patterns using reduced rank regression and to examine their relations to depressive symptoms. Participants were 2006 Japanese employees aged 19–69 years. Depressive symptoms were assessed using the Center for Epidemiologic Studies Depression Scale. Diet was assessed using a validated, self-administered diet history questionnaire. Dietary patterns were extracted by reduced rank regression with 6 depression-related nutrients as response variables. Logistic regression was used to estimate odds ratios of depressive symptoms adjusted for potential confounders. A dietary pattern characterized by a high intake of vegetables, mushrooms, seaweeds, soybean products, green tea, potatoes, fruits, and small fish with bones and a low intake of rice was associated with fewer depressive symptoms. The multivariable-adjusted odds ratios of having depressive symptoms were 0.62 (95% confidence interval, 0.48–0.81) in the highest versus lowest tertiles of dietary score. Results suggest that adherence to a diet rich in vegetables, fruits, and typical Japanese foods, including mushrooms, seaweeds, soybean products, and green tea, is associated with a lower probability of having depressive symptoms. & 2015 Published by Elsevier Ireland Ltd.
Keywords: Depression Dietary pattern Epidemiology Japanese Reduced rank regression
1. Introduction Depression, a common condition in the general population, reduces work productivity, lowers quality of life, and increases mortality (Doris et al., 1999). Previous epidemiologic evidence suggested that several kinds of nutrients and foods, such as n-3 polyunsaturated fatty acids, folate, other vitamins, minerals, fruits, and vegetables, are beneficial against depression, but the results have been generally inconclusive (Murakami and Sasaki, 2010). In addition, the investigation of foods or nutrients in isolation may not be a suitable way to assess the interactive or synergic effects n Corresponding author at: Department of Mental Health, Graduate School of Medicine, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan. Fax: þ81 3 5841 3592. E-mail address: [email protected] (T. Miki).
among nutrients (Hu, 2002). In daily life, we do not consume foods or nutrients in isolation but rather in combination as meals. This fact highlights the importance of exploring dietary patterns associated with depression. Dietary pattern analysis, which assesses the effect of overall diet on disease risk, has emerged as a complement to traditional approaches (Hu, 2002). This method may be particularly useful when multiple nutrients are involved in the development of a disease (Hu, 2002). Two types of technique have been commonly used to examine the association between dietary pattern and depression: a priori defined methods (e.g., diet quality score) and exploratory methods (e.g., principal component analysis). In their recent meta-analysis of dietary patterns derived by these methods, Lai et al. suggested that a healthy diet was associated with reduced odds of depression. Besides, the authors also found significant heterogeneity among studies, making it difficult to obtain a
http://dx.doi.org/10.1016/j.psychres.2015.07.033 0165-1781/& 2015 Published by Elsevier Ireland Ltd.
Please cite this article as: Miki, T., et al., Dietary patterns derived by reduced rank regression (RRR) and depressive symptoms in Japanese employees: The Furukawa nutrition and health study. Psychiatry Research (2015), http://dx.doi.org/10.1016/j. psychres.2015.07.033i
T. Miki et al. / Psychiatry Research ∎ (∎∎∎∎) ∎∎∎–∎∎∎
2
reliable summary estimate of the association between dietary patterns and depression (Lai et al., 2014). These methods have advantages and limitations. Diet quality score, which is calculated based on existing knowledge such as recommended diet, captures only limited aspects of the diet and does not consider the correlation structure of food and nutrient intakes (Hoffmann et al., 2004). In contrast, an exploratory statistical approach determines dietary patterns based solely on inter-correlations among dietary variables. However, this method does not incorporate any existing knowledge of nutrient-disease associations and may not capture dietary patterns relevant to disease risk (Hoffmann et al., 2004). Neither method aims to identify dietary patterns that are specific to a disease of interest (Michels and Schulze, 2005). To overcome these issues, a new a posteriori method has been proposed, reduced rank regression (RRR) (Hoffmann et al., 2004). This technique identifies dietary patterns based on several nutrients or biomarkers that have been linked to the disease of interest. This combination of disease-specific knowledge and dietary information of the study population suggests that RRR is an appropriate and promising statistical method to determine dietary patterns associated with an outcome of interest (Hoffmann et al., 2004). To date, RRR has been applied to explore the relationship between an inflammation-prone dietary pattern and depression risk (Lucas et al., 2014). To our knowledge, however, it has not been used to identify dietary patterns based on knowledge of nutrient-mood associations. Here, we used RRR to identify dietary patterns and investigate their association with depressive symptoms among Japanese employees.
2. Methods 2.1. Study procedure and participants Data for the present study was derived from the Furukawa Nutrition and Health Study, which has been described in detail elsewhere (Nanri et al., 2014). A nutritional epidemiological survey was conducted during periodic health checks in April 2012 (factory A) and May 2013 (factory B) among employees of a manufacturing company and their affiliated companies located in Japan's Kanto region. Of 2828 health examination attendants (11% women), 2162 employees agreed to participate in the study (response rate, 76%). We assessed lifestyle, including diet and health status, and obtained health examination data. The protocol of the study was approved by the ethics committee of the National Center for Global Health and Medicine, Japan, and secondary analysis of the Furukawa Nutrition and Health Study data was approved by the ethics committee of the University of Tokyo. Written informed consent was obtained from each participant. We excluded 100 participants with a history of cancer (n ¼20); cardiovascular disease (n¼ 25); chronic hepatitis (n ¼2); kidney disease, including nephritis (n ¼11); pancreatitis (n ¼3); and mental disorders, such as depression and neurotic disorder (n ¼45). Some participants had two or more these conditions. We excluded these participants because such diseases might affect dietary habits or depressive symptoms and thereby cause reverse causality. Of the remaining 2062, we excluded 11 individuals who did not return the study questionnaires and 33 participants who had missing data on the outcome and covariates of the present analysis. We additionally excluded 12 participants with extreme total energy intake ( 43 standard deviations). Finally, 2006 participants (1792 men and 214 women) aged 19–69 years old were included in the analysis.
2.2. Depressive symptoms Depressive symptoms were assessed using a Japanese version (Shima et al., 1985) of the Center for Epidemiologic Studies Depression (CES-D) scale (Radloff, 1977). This scale consists of 20 items addressing 6 typical symptoms of depression experienced during the preceding week, including depressed mood, guilt or worthlessness, helplessness or hopelessness, psychomotor retardation, loss of appetite, and sleep disturbance. Each item is scored on a scale of 0–3 according to the frequency of the symptom, and the scores are then summed to give the total CES-D score, ranging from 0 to 60. The criterion validity of the CES-D scale has been well established in both Western (Radloff, 1977) and Japanese (Shima et al., 1985) subjects. Participants with a CES-D score Z 16 were considered to have depressive symptoms (Radloff, 1977). Another cutoff of Z 19, which might be suitable for Japanese workers (Wada et al., 2007), was also used. 2.3. Dietary assessment Dietary habits during the preceding one-month period were assessed using a validated brief self-administered diet history questionnaire (BDHQ) (Kobayashi et al., 2012) consisting of five sections: (1) intake frequency of 46 food and non-alcoholic beverage items; (2) daily intake of rice and miso soup; (3) frequency of alcoholic drinking and amount of consumption of five alcoholic beverages per typical drinking occasion; (4) usual cooking method; and (5) general dietary behavior. Dietary intakes for 58 food and beverage items, energy, and selected nutrients were estimated using an ad hoc computer algorithm for the BDHQ (Kobayashi et al., 2011), with reference to the Standard Tables of Food Composition in Japan (Science and Technology Agency, 2005a; Science and Technology Agency, 2010b). According to a validation study of the BDHQ using 16-day weighted dietary records as standard, correlation coefficients were 40.40 for the intake of many foods, beverages (Kobayashi et al., 2011), and nutrients used in our study (Kobayashi et al., 2012). 2.4. Other variables The survey questionnaire assessed marital status, job grade, night and rotating shift work, overtime work, smoking, physical activity during work and housework or in commuting to work, and leisure-time physical activity. Physical activity during work and housework or in commuting and leisure-time were expressed as the sum of metabolic equivalents (METs) multiplied by the duration of time (in hours) across physical activities with different levels. 2.5. Statistical analysis Dietary patterns were determined using RRR techniques (Hoffmann et al., 2004). RRR identifies linear functions of food groups (i.e., the dietary patterns) that explain as much of the variation of selected nutrients that are potentially protective or contributing factors to the relevant disease as possible. Therefore, the score for dietary patterns analyzed by this method is likely to be associated with the disease. We selected the following six nutrients as potentially protective factors for depression: folate (Gilbody et al., 2007), vitamin C (Woo et al., 2006; Oishi et al., 2009), magnesium (Jacka et al., 2009, 2012; Yary et al., 2013), calcium (Bae and Kim, 2012), iron (Woo et al., 2006), and zinc (Amani et al., 2010; Jacka et al., 2012; Maserejian et al., 2012; Yary and Aazami, 2012; Vashum et al., 2014). In addition, we found that these nutrients were inversely associated with depressive symptoms in a preliminary analysis in our study population. Dietary
Please cite this article as: Miki, T., et al., Dietary patterns derived by reduced rank regression (RRR) and depressive symptoms in Japanese employees: The Furukawa nutrition and health study. Psychiatry Research (2015), http://dx.doi.org/10.1016/j. psychres.2015.07.033i
T. Miki et al. / Psychiatry Research ∎ (∎∎∎∎) ∎∎∎–∎∎∎
3
Table 1 Explained percentage of variations in nutrients (response variables) and food and beverage items by extracted dietary patterns.a
Explained variation in nutrients (%) Explained variation in food and beverage items (%) Pearson correlation coefficients b Folate Vitamin C Magnesium Calcium Iron Zinc
DP1
DP2
DP3
DP4
DP5
DP6
Total explained variation
73.2 8.6
11.5 2.4
7.3 2.1
4.4 2.8
2.3 1.8
1.1 2.2
99.9 19.9
0.92c 0.84c 0.90c 0.80c 0.94c 0.71c
0.29c 0.40c 0.02 0.21c 0.003 0.63c
0.12c 0.10c 0.17c 0.52c 0.19c 0.29c
0.06 0.32c 0.33c 0.16c 0.13c 0.11c
0.19c 0.16c 0.21c 0.10c 0.13c 0.06
0.16c 0.03 0.03 0.01 0.18c 0.06
Abbreviation: DP, Dietary Pattern. a b c
Dietary patterns were derived by reduced rank regression analysis. Pearson correlation coefficients between nutrients (response variables) and extracted dietary patterns. Po 0.001.
patterns related to the intake of these 6 nutrients were derived on the basis of 52 food and beverage items, excluding 6 items (sugar added to coffee and black tea, three items usually added during cooking [salt, oil, and sugar], table salt and salt-containing seasoning at the table, and soup consumed with noodles). The 6 nutrients selected as response variables and 52 food and beverage items were energy-adjusted prior to the RRR using the density method. Although n-3 polyunsaturated fatty acid levels have been shown to be inversely associated with depressive symptoms (Murakami and Sasaki, 2010), the present and previous studies among Japanese (Miyake et al., 2006; Murakami et al., 2008) found no significant association between them. Accordingly, we decided not to include n-3 polyunsaturated fatty acids as response variables. Explained percentage of variations in nutrients (response variables) and food and beverage items by extracted dietary patterns and Pearson correlation coefficients between nutrients and dietary patterns were calculated. Of the six dietary patterns extracted, we focused on dietary pattern 1 (DP1) in our statistical analysis because it explains the largest amount of variation among the nutrients. Moreover, when dietary patterns were extracted on the basis of a reduced set of food groups (22 food groups), similar dietary patterns were observed. We calculated factor loadings, which represent the magnitude and direction of each food and beverage item's contribution to DP1 scores, of food and beverage items associated with DP1 and Pearson correlation coefficients between food and beverage items and nutrients (response variables). A positive factor loading value indicated that a higher DP1 score was associated with an increased intake for that food or beverage items. On the other hand, a negative factor loading value indicated less intake of the items. To show the general characteristics of the study population, the mean or proportion was presented according to the tertile of DP1 score. Trend association across tertile categories of scores for DP1 was assessed using the Mantel–Haenszel χ2-test for categorical variables and linear regression analysis for continuous variables, assigning ordinal numbers 1–3 to tertile categories of DP1 scores. To examine the association between DP1 and depressive symptoms, we performed a multiple logistic regression analysis and calculated the odds ratios (ORs) and 95% confidence intervals (CIs) of depressive symptoms for the tertiles of DP1 scores, using the lowest tertile category as reference. The first model was adjusted for age (in years, continuous), sex, and site (survey in April 2012 or in May 2013). The second model was further adjusted for marital status (married or other), job grade (low, middle, or high), night or rotating shift work (yes or no), overtime work (o10 h/month, 10– o30 h/month, or Z30 h/month), physical activity at work and housework or in commuting to work (o3 METs-h/day, 3– o7 METs-h/day, 7– o20 METs-h/day, or Z20 METs-h/day),
leisure-time physical activity (not engaged, 4 0– o3 METs-h/ week, 3– o10 METs-h/week, or Z10 METs-h/week), smoking (never-smoker, quitter, current smoker consuming o20 cigarettes/day, or current smoker consuming Z20 cigarettes/day), and total energy intake (kcal/day, continuous). We repeated the analysis after excluding participants with obesity, a history of diabetes, and/or hypertension, on the basis that these conditions might affect dietary habits and may lead to reverse causality. Additionally, we randomly divided the sample in half and analyzed the data in two subsamples to confirm that the results were reproducible in subsamples. Two-sided P o0.05 was considered statistically significant in all analyses. The RRR analyses were performed using Statistical Analysis System (SAS) software version 9.4 (SAS Institute, Cary, NC, USA), while all the other analyses were performed using Stata version 12.1 (StataCorp, College Station, TX, USA).
3. Results DP1 scores were highly correlated with each nutrient (folate, vitamin C, magnesium, calcium, iron, and zinc). All Pearson correlation coefficients were greater than 0.70 (Table 1). Factor loadings of the 52 food and beverage items associated with DP1 scores and correlation coefficients between food and beverage items and the 6 response variables are presented in Table 2. Food items with a factor loading greater than 0.15 were a high intake of vegetables, mushrooms, seaweeds, soybean products, green tea, potatoes, fruits, and small fish with bones. In contrast, the only food item with a factor loading less than –0.15 was rice. Table 3 shows the characteristics of study participants according to tertile of DP1 score. Participants with higher DP1 scores were older, more likely to be women and physically active in leisure, but less likely to be night or rotating shift workers, overtime workers, smokers, and alcohol drinkers. Mean values of folate, vitamin C, magnesium, calcium, iron, and zinc increased with higher DP1 score. The prevalence of depressive symptoms was 27.8%. The ORs of depressive symptoms according to tertile categories of DP1 are shown in Table 4. In an age-, sex-, and workplace-adjusted model (model 1), DP1 was significantly associated with a decreased prevalence of depressive symptoms. Further adjustment for other covariates (model 2) somewhat attenuated the association, but it nevertheless remained significant. The multivariate-adjusted ORs for the lowest through highest tertile of DP1 were 1.00 (reference), 0.78 (95% CI, 0.61–0.99), and 0.62 (95% CI, 0.48–0.81), respectively (P for trend o0.001). When a higher cutoff value for the definition of depressive symptoms was used (CES-D Z19), the ORs for depressive symptoms were 1.00 (reference), 0.93 (95% CI, 0.70–1.23),
Please cite this article as: Miki, T., et al., Dietary patterns derived by reduced rank regression (RRR) and depressive symptoms in Japanese employees: The Furukawa nutrition and health study. Psychiatry Research (2015), http://dx.doi.org/10.1016/j. psychres.2015.07.033i
T. Miki et al. / Psychiatry Research ∎ (∎∎∎∎) ∎∎∎–∎∎∎
4
Table 2 Factor loadings of food and beverage items associated with dietary pattern 1 and correlation coefficients between food and beverage items and nutrients (response variables)a. Food and beverage items
Low-fat milk and yogurt Milk and yogurt Squid/octopus/shrimp/shellfish Small fish with bones Dried fish/salted fish Oily fish Lean fish Egg Tofu/atsuage b Natto c Potatoes Pickled green leaves vegetables Lettuces/cabbage (raw) Green leaves vegetables Cabbage/Chinese cabbage Carrots/pumpkin Japanese radish/turnip Other root vegetables Tomatoes Mushrooms Seaweeds Citrus fruit Persimmons/strawberries/kiwifruit Other fruit Green tea Cola drink/soft drink Rice
Factor loadings
Correlation between food items and 6 response variables
Dietary pattern 1
Folate
Vitamin C
Magnesium
Calcium
Iron
Zinc
0.11 0.11 0.12 0.15 0.14 0.12 0.14 0.13 0.21 0.17 0.16 0.13 0.23 0.33 0.27 0.25 0.21 0.24 0.19 0.24 0.20 0.15 0.14 0.15 0.17 0.11 0.15
0.11 0.12 0.19 0.22 0.19 0.18 0.23 0.23 0.34 0.29 0.30 0.27 0.53 0.76 0.62 0.54 0.43 0.52 0.41 0.48 0.37 0.28 0.25 0.26 0.45 0.21 0.30
0.11 0.16 0.14 0.16 0.15 0.15 0.22 0.14 0.25 0.15 0.38 0.22 0.47 0.65 0.57 0.53 0.40 0.50 0.43 0.44 0.33 0.46 0.47 0.41 0.43 0.16 0.28
0.21 0.15 0.29 0.38 0.34 0.28 0.34 0.18 0.50 0.46 0.34 0.27 0.42 0.61 0.48 0.49 0.42 0.47 0.35 0.47 0.42 0.26 0.22 0.28 0.22 0.24 0.35
0.53 0.49 0.15 0.39 0.30 0.22 0.23 0.20 0.39 0.23 0.20 0.23 0.38 0.53 0.41 0.41 0.33 0.35 0.32 0.37 0.33 0.27 0.26 0.32 0.16 0.16 0.34
0.12 0.07 0.25 0.27 0.36 0.29 0.30 0.36 0.47 0.41 0.32 0.29 0.44 0.67 0.49 0.46 0.39 0.45 0.35 0.46 0.40 0.24 0.22 0.23 0.39 0.22 0.31
0.21 0.21 0.34 0.29 0.23 0.23 0.25 0.32 0.36 0.29 0.23 0.16 0.28 0.37 0.29 0.31 0.25 0.30 0.25 0.33 0.27 0.16 0.15 0.15 0.14 0.22 0.02
a Omitted in the tables were food items with factor loadings less than 70.10 for dietary pattern 1 (chicken, pork/beef, ham/sausage/bacon, liver, canned tuna, other pickles, western-type confectioneries, Japanese confectioneries, rice crackers/rice cake/okonomiyaki: meat/fish and vegetables pancake, ice cream, mayonnaise/dressing, bread, buckwheat noodles, Japanese wheat noodles, Chinese noodles, spaghetti and macaroni, black tea/oolong tea, coffee, 100% fruit and vegetable juice, miso soup, sake, beer, shochu, whisky, and wine). b Deep-fried tofu. c Fermented soybeans.
and 0.69 (95% CI, 0.51–0.94), respectively (P for trend¼0.019). A stronger association was observed after excluding participants
with obesity, a history of diabetes, and/or hypertension (n ¼1407); the adjusted ORs of depressive symptoms for the lowest through
Table 3 Characteristics of the study population by tertiles of dietary pattern 1 scoresa. Variables
Number of subjects Age (mean7 s.d., year) Sex (women, %) Site (survey in April 2012, %) Marital status (married, %) Job grade (low, %) Night or rotating shift work (yes, %) Overtime work ( Z 30 h/month, %) Physical activity at work and housework or in commuting to work (Z 20 METs-h/day, %) Leisure-time physical activities ( Z10 METs-h/week, %) Smoking status (current, %) Alcohol drinking (currentb, %) Daily dietary intake (mean7 s.d) Total energy (kcal/day) Folate (μg /1000 kcal) Vitamin C (mg/1000 kcal) Magnesium (mg/1000 kcal) Calcium (mg/1000 kcal) Iron (mg/1000 kcal) Zinc (mg/1000 kcal)
Dietary pattern 1 scores Tertile 1 (low)
Tertile 2
Tertile 3 (high)
P-trenda
669 40.5 7 9.3 5.8 52.0 62.6 71.5 31.5 28.4 30.8 20.8 39.8 57.4
669 42.17 9.9 9.9 55.5 68.8 68.3 17.9 27.2 20.5 27.7 25.6 55.9
668 44.0 7 10.7 16.3 59.9 66.6 67.4 10.9 24.1 17.7 29.9 21.1 45.7
o 0.001 o 0.001 0.015 0.056 0.14 o 0.001 o 0.001 o 0.001 o 0.001 o 0.001 o 0.001
1785 7 511 1137 26 317 12 102 7 14 1657 49 2.9 7 0.5 3.7 7 0.5
1836 7 467 1567 25 48 7 12 1237 12 2277 52 3.7 70.4 4.2 70.4
1768 7488 222 7 60 727 25 1527 23 3107 84 4.8 7 0.9 4.6 7 0.5
0.69 o 0.001 o 0.001 o 0.001 o 0.001 o 0.001 o 0.001
Abbreviations: s.d., standard deviation; METs; Metabolic Equivalents a On the basis of the Mantel–Haenszel χ2-test for categorical variables and linear regression analysis for continuous variables, assigning ordinal numbers 1–3 to tertile categories of scores for dietary pattern 1. b Alcohol consumption of at least one day per week.
Please cite this article as: Miki, T., et al., Dietary patterns derived by reduced rank regression (RRR) and depressive symptoms in Japanese employees: The Furukawa nutrition and health study. Psychiatry Research (2015), http://dx.doi.org/10.1016/j. psychres.2015.07.033i
T. Miki et al. / Psychiatry Research ∎ (∎∎∎∎) ∎∎∎–∎∎∎ Table 4 Odds ratios and 95% confidence intervals for depressive symptoms according to tertiles of dietary pattern 1 scores. Variables
CES-D (15/16) Subjects with/without depressive symptoms Model 1b Model 2c CES-D (18/19) Subjects with/without depressive symptoms Model 1b Model 2c
Dietary pattern 1 scores Tertile 1 (low)
Tertile 2
Tertile 3 (high)
234/435
180/489
143/525
1.00 (ref)
0.70 (0.56– 0.89) 0.78 (0.61– 0.99)
0.54 (0.42– 0.69) 0.62 (0.48– 0.81)
150/519
124/545
89/579
1.00 (ref)
0.82 (0.63– 1.07) 0.93 (0.70– 1.23)
0.58 (0.44– 0.78) 0.69 (0.50– 0.94)
1.00 (ref)
1.00 (ref)
P-trenda
o 0.001 o 0.001
o 0.001 0.019
Abbreviations: CES-D, Center for Epidemiologic Studies Depression Scale; ref, reference. a Based on multiple logistic regression analyses, assigning ordinal numbers of 1–3 to the tertile categories of scores for dietary pattern 1. b Adjusted for age (year, continuous), sex, and site (survey in April 2012 or in May 2013). c Adjusted for age (year, continuous), sex, site (survey in April 2012 or in May 2013), marital status (married or other), job grade (low, middle, or high), night or rotating shift work (yes or no), overtime work ( o 10 h/month, 10– o30 h/month, or Z 30 h/month), physical activity at work and housework or in commuting to work ( o 3 METs-h/day, 3– o7 METs-h/day, 7–o 20 METs-h/day, or Z 20 METs-h/ day), leisure-time physical activity (not engaged, 4 0–o 3 METs-h/week, 3–o 10 METs-h/week, or Z 10 METs-h/week), smoking (never-smoker, quitter, current smoker consuming o20 cigarettes/day, or current smoker consuming Z20 cigarettes/day), and total energy intake (kcal/day, continuous).
highest tertile of DP1 in model 2 was 1.00 (reference), 0.63 (95% CI, 0.46–0.84), and 0.50 (95% CI, 0.37–0.69), respectively (P for trend o0.001). In addition, inverse associations were observed in both randomly divided subsamples.
4. Discussion In this study in Japanese workers, RRR identified a dietary pattern that was associated with a lower prevalence of depressive symptoms. This dietary pattern was characterized by a high intake of vegetables, mushrooms, seaweeds, soybean products, green tea, potatoes, fruits, and small fish with bones, as well as a low intake of rice. To our knowledge, this is the first study to investigate the association between dietary pattern based on existing knowledge of nutrients and depressive symptoms using an empirical dietary pattern method, RRR. Dietary patterns similar to the present one (DP1) have been linked to depressive symptoms. In a recent meta-analysis of observational studies that used a priori methods (e.g., diet quality scores or indexes) and a posteriori approaches (e.g., factor analysis, including principal component analysis) in the identification of dietary patterns, a healthy diet pattern, characterized by high intake of fruits, vegetables, fish, and whole grains, which is similar to our present derived dietary pattern, was inversely associated with depression (Lai et al., 2014). The present findings are also consistent with those of two other Japanese studies that identified dietary patterns using principal component analysis: a healthy Japanese dietary pattern characterized by a high intake of vegetables, fruit, mushrooms, and soy products was associated with a decreased prevalence of depressive symptoms (Nanri et al., 2010); and a balanced Japanese dietary pattern featuring a high intake of
5
vegetables, mushrooms, and seaweed was related to a decreased prevalence of depressive symptoms (Suzuki et al., 2013). Even if a dietary pattern extracted using principal component analysis is found to be associated with a specific disease or condition, the interpretation of such findings is often difficult due to their exploratory nature (Hoffmann et al., 2004). In contrast, the association with a dietary pattern derived using an empirical RRR method may provide some insight into mediators linking food intake and disease. Although the dietary pattern in the present study was similar to those identified using principal component analysis (Nanri et al., 2010; Suzuki et al., 2013), our results provide additional evidence that a diet rich in potentially important nutrients (folate, vitamin C, magnesium, calcium, iron, and zinc) may be inversely associated with depressive symptoms. The inverse association between DP1 and depressive symptoms in the present study is well supported by the results of epidemiological studies on the association of depression with folate (Gilbody et al., 2007), vitamin C (Woo et al., 2006; Oishi et al., 2009), magnesium (Jacka et al., 2009, 2012; Yary et al., 2013), calcium (Bae and Kim, 2012), iron (Woo et al., 2006), and zinc (Amani et al., 2010; Jacka et al., 2012; Maserejian et al., 2012; Yary and Aazami, 2012; Vashum et al., 2014). These nutrients may independently as well as jointly protect against depression through multiple pathways. Folate is involved in the metabolism of monoamines in the brain, such as serotonin, and may protect brain function via its ability to reduce homocysteine, which has a neurotoxic effect (Bottiglieri, 2005). Vitamin C (ascorbic acid) is required to recycle tetrahydrobiopterin, which is necessary for tryptophan hydroxylase activity in serotonin synthesis (Meredith and May, 2013). In addition, reactive oxygen species (ROS) and defective antioxidant defenses are thought to be involved in the pathophysiology of depression (Bilici et al., 2001). Thus, antioxidant vitamins, such as vitamin C, may prevent depression. Magnesium is necessary for tryptophan hydroxylase and serotonin receptor binding (Kantak, 1988) and is a potent antagonist of the N-methyl-D-aspartate (NMDA) receptor complex (Bresink et al., 1995). Calcium activates tryptophan hydroxylase in the synthesis of serotonin (Knapp et al., 1975). Iron is involved in the synthesis of neurotransmitters, including dopamine and serotonin (Beard et al., 1993). Zinc might modulate mood by enhancing the function of the serotonergic system or inhibiting the function of the NMDA receptor complex in the central nervous system (Szewczyk et al., 2008). The major strengths of this study include its high participation rate and use of validated methodologies to assess nutrient intake (i.e., the BDHQ). Some limitations should also be noted. First, an association derived from a cross-sectional study design does not necessarily indicate causality. For instance, the lower intake of food among persons with depressed mood than those without may be caused by decreased appetite due to the depressive state. However, this is less likely in our study because total energy intake between participants with and without depressed mood was similar (1827 kcal and 1785 kcal, respectively). Nevertheless, we could not exclude the possibility that the choice of foods determining the dietary pattern had changed as a result of the depressive state. Second, we chose depression-related nutrients as response variables, rather than biomarkers, in RRR. Given that the dietary intake of a nutrient does not necessary reflect body nutritional status, which is also determined by absorption, metabolism, and excretion, an RRR study that includes nutritional biomarkers as response variables is warranted. Third, we assessed depressive symptoms using a validated questionnaire (i.e., the CES-D scale), but not structured diagnostic interviews. The association may differ if the outcome is clinically diagnosed depression. Fourth, although we adjusted for a range of potential confounding variables, we cannot rule out the possibility that the
Please cite this article as: Miki, T., et al., Dietary patterns derived by reduced rank regression (RRR) and depressive symptoms in Japanese employees: The Furukawa nutrition and health study. Psychiatry Research (2015), http://dx.doi.org/10.1016/j. psychres.2015.07.033i
T. Miki et al. / Psychiatry Research ∎ (∎∎∎∎) ∎∎∎–∎∎∎
6
observed associations were due to unmeasured confounders and residual confounding. Finally, because the present study was conducted among workers in a Japanese manufacturing company, caution is required in generalizing the findings. In conclusion, this study suggests that adherence to a diet rich in vegetables, fruits, and typical Japanese foods, including mushrooms, seaweeds, soybean products, and green tea, may be associated with a lower prevalence of depressive symptoms. The observed association in this cross-sectional study requires confirmation in prospective studies among populations with different backgrounds.
Contributors T. Mizoue and A.N. designed the research; T.K., M.E., K. Kuwahara, H.T., K. Kurotani, R. I., I. Kabe, T. Mizoue, and A.N. conducted the research; and T. Miki performed the statistical analysis, wrote the manuscript, and had primary responsibility for the final content. All authors were involved in revision of the manuscript and approved the final version of the manuscript. None of the authors had any conflicts of interest. T.K., M.E., H.T., R.I., and I. Kabe are health professionals in Furukawa Electric Corporation.
Conflict of interest The authors declare no conflict of interest. T.K., M.E., H.T., R.I., and I. Kabe are health professionals in Furukawa Electric Corporation.
Acknowledgments We thank Fumiko Zaizen (Furukawa Electric Corporation) and Ayami Kume, Sachiko Nishihara, Yuho Mizoue, Saeko Takagiwa, and Yuriko Yagi (National Center for Global Health and Medicine) for their help in data collection. This study was supported by a Grant-in-Aid for Young Scientists (A) (25702006) from the Japan Society for the Promotion of Science, Health and Labour Sciences Research Grant, and Health and Labour Sciences Research Expenses for Commission (Comprehensive Research on Life-Style Related Diseases including Cardiovascular Diseases and Diabetes Mellitus H26-005).
References Amani, R., Saeidi, S., Nazari, Z., Nematpour, S., 2010. Correlation between dietary zinc intakes and its serum levels with depression scales in young female students. Biol. Trace Elem. Res. 137, 150–158. Bae, Y.J., Kim, S.K., 2012. Low dietary calcium is associated with self-rated depression in middle-aged Korean women. Nutr. Res. Pract. 6, 527–533. Beard, J.L., Connor, J.R., Jones, B.C., 1993. Iron in the brain. Nutrition Reviews 51, 157–170. Bilici, M., Efe, H., Koroglu, M.A., Uydu, H.A., Bekaroglu, M., Deger, O., 2001. Antioxidative enzyme activities and lipid peroxidation in major depression: alterations by antidepressant treatments. J. Affect. Disord. 64, 43–51. Bottiglieri, T., 2005. Homocysteine and folate metabolism in depression. Prog. Neuro-Psychopharmacol. Biol. Psychiatry 29, 1103–1112. Bresink, I., Danysz, W., Parsons, C.G., Mutschler, E., 1995. Different binding affinities of NMDA receptor channel blockers in various brain regions—indication of NMDA receptor heterogeneity. Neuropharmacology 34, 533–540. Doris, A., Ebmeier, K., Shajahan, P., 1999. Depressive illness. Lancet 354, 1369–1375. Gilbody, S., Lightfoot, T., Sheldon, T., 2007. Is low folate a risk factor for depression? A meta-analysis and exploration of heterogeneity. J. Epidemiol. Commun. Health 61, 631–637. Hoffmann, K., Schulze, M.B., Schienkiewitz, A., Nothlings, U., Boeing, H., 2004. Application of a new statistical method to derive dietary patterns in nutritional epidemiology. Am. J. Epidemiol. 159, 935–944. Hu, F.B., 2002. Dietary pattern analysis: a new direction in nutritional
epidemiology. Curr. Opin. Lipidol. 13, 3–9. Jacka, F.N., Overland, S., Stewart, R., Tell, G.S., Bjelland, I., Mykletun, A., 2009. Association between magnesium intake and depression and anxiety in community-dwelling adults: the Hordaland Health Study. Aust. N. Z. J. Psychiatry 43, 45–52. Jacka, F.N., Maes, M., Pasco, J.A., Williams, L.J., Berk, M., 2012. Nutrient intakes and the common mental disorders in women. J. Affect. Disord. 141, 79–85. Kantak, K.M., 1988. Magnesium deficiency alters aggressive behavior and catecholamine function. Behav. Neurosci. 102, 304–311. Knapp, S., Mandell, A.J., Bullard, W.P., 1975. Calcium activation of brain tryptophan hydroxylase. Life Sci. 16, 1583–1593. Kobayashi, S., Murakami, K., Sasaki, S., Okubo, H., Hirota, N., Notsu, A., Fukui, M., Date, C., 2011. Comparison of relative validity of food group intakes estimated by comprehensive and brief-type self-administered diet history questionnaires against 16 d dietary records in Japanese adults. Public Health Nutr. 14, 1200–1211. Kobayashi, S., Honda, S., Murakami, K., Sasaki, S., Okubo, H., Hirota, N., Notsu, A., Fukui, M., Date, C., 2012. Both comprehensive and brief self-administered diet history questionnaires satisfactorily rank nutrient intakes in Japanese adults. J. Epidemiol. 22, 151–159. Lai, J.S., Hiles, S., Bisquera, A., Hure, A.J., McEvoy, M., Attia, J., 2014. A systematic review and meta-analysis of dietary patterns and depression in communitydwelling adults. Am. J. Clin. Nutr. 99, 181–197. Lucas, M., Chocano-Bedoya, P., Shulze, M.B., Mirzaei, F., O’Reilly, E.J., Okereke, O.I., Hu, F.B., Willett, W.C., Ascherio, A., 2014. Inflammatory dietary pattern and risk of depression among women. Brain Behav. Immun. 36, 46–53. Maserejian, N.N., Hall, S.A., McKinlay, J.B., 2012. Low dietary or supplemental zinc is associated with depression symptoms among women, but not men, in a population-based epidemiological survey. J. Affect. Disord. 136, 781–788. Meredith, M.E., May, J.M., 2013. Regulation of embryonic neurotransmitter and tyrosine hydroxylase protein levels by ascorbic acid. Brain Res. 1539, 7–14. Michels, K.B., Schulze, M.B., 2005. Can dietary patterns help us detect diet-disease associations? Nutr. Res. Rev. 18, 241–248. Miyake, Y., Sasaki, S., Yokoyama, T., Tanaka, K., Ohya, Y., Fukushima, W., Saito, K., Ohfuji, S., Kiyohara, C., Hirota, Y., 2006. Risk of postpartum depression in relation to dietary fish and fat intake in Japan: the Osaka Maternal and Child Health Study. Psychol. Med. 36, 1727–1735. Murakami, K., Mizoue, T., Sasaki, S., Ohta, M., Sato, M., Matsushita, Y., Mishima, N., 2008. Dietary intake of folate, other B vitamins, and omega-3 polyunsaturated fatty acids in relation to depressive symptoms in Japanese adults. Nutrition 24, 140–147. Murakami, K., Sasaki, S., 2010. Dietary intake and depressive symptoms: a systematic review of observational studies. Mol. Nutr. Food Res. 54, 471–488. 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. Nanri, A., Eguchi, M., Kuwahara, K., Kochi, T., Kurotani, K., Ito, R., Pham, N.M., Tsuruoka, H., Akter, S., Jacka, F., Mizoue, T., Kabe, I., 2014. Macronutrient intake and depressive symptoms among Japanese male workers: the Furukawa Nutrition and Health Study. Psychiatry Res. 220, 263–268. Oishi, J., Doi, H., Kawakami, N., 2009. Nutrition and depressive symptoms in community-dwelling elderly persons in Japan. Acta Med. Okayama 63, 9–17. Radloff, L.S., 1977. The CES-D scale: a self-report depression scale for research in the general population. Appl. Psychol. Meas. 1, 385–401. Science and Technology Agency, 2005a. Standard Tables of Food Composition in Japan, fatty acids section. 5th revised and enlarged ed. Printing Bureau of the Ministry of Finance, Tokyo. (in Japanese). Science and Technology Agency, 2010b. Standard Tables of Food Composition in Japan, 2010. Tokyo: Ministry of Education, Culture, Sports, Science and Technology. (in Japanese). Shima, S., Shikano, T., Kitamura, T., Asai, M., 1985. New self-rating scale for depression. Japanese. J. Clin. Psychiatry 27, 717–723. Suzuki, T., Miyaki, K., Tsutsumi, A., Hashimoto, H., Kawakami, N., Takahashi, M., Shimazu, A., Inoue, A., Kurioka, S., Kakehashi, M., Sasaki, Y., Shimbo, T., 2013. Japanese dietary pattern consistently relates to low depressive symptoms and it is modified by job strain and worksite supports. J. Affect. Disord. 150, 490–498. Szewczyk, B., Poleszak, E., Sowa-Kucma, M., Siwek, M., Dudek, D., Ryszewska-Pokrasniewicz, B., Radziwon-Zaleska, M., Opoka, W., Czekaj, J., Pilc, A., Nowak, G., 2008. Antidepressant activity of zinc and magnesium in view of the current hypotheses of antidepressant action. Pharmacol. Rep. 60, 588–589. Vashum, K.P., McEvoy, M., Milton, A.H., McElduff, P., Hure, A., Byles, J., Attia, J., 2014. Dietary zinc is associated with a lower incidence of depression: findings from two Australian cohorts. J. Affect. Disord. 166, 249–257. Wada, K., Tanaka, K., Theriault, G., Satoh, T., Mimura, M., Miyaoka, H., Aizawa, Y., 2007. Validity of the Center for Epidemiologic Studies Depression Scale as a screening instrument of major depressive disorder among Japanese workers. Am. J. Ind. Med. 50, 8–12. Woo, J., Lynn, H., Lau, W.Y., Leung, J., Lau, E., Wong, S.Y., Kwok, T., 2006. Nutrient intake and psychological health in an elderly Chinese population. Int. J. Geriatr. Psychiatry 21, 1036–1043. Yary, T., Aazami, S., 2012. Dietary intake of zinc was inversely associated with depression. Biol. Trace Elem. Res. 145, 286–290. Yary, T., Aazami, S., Soleimannejad, K., 2013. Dietary intake of magnesium may modulate depression. Biol. Trace Elem. Res. 151, 324–329.
Please cite this article as: Miki, T., et al., Dietary patterns derived by reduced rank regression (RRR) and depressive symptoms in Japanese employees: The Furukawa nutrition and health study. Psychiatry Research (2015), http://dx.doi.org/10.1016/j. psychres.2015.07.033i