Introduction

Depression is a common mental illness and the leading major contributor to the overall global burden of disease worldwide, with an estimated 350 million people affected [1]. There is evidence that depressive symptoms are associated with myocardial infarction and diabetes [2]. While the mechanisms of depressive symptoms are largely unknown, the importance of research into the possible role of dietary factors is emphasized by the fact that diet is modifiable. Attention has recently focused on dietary glycemic index (GI) [3] and glycemic load (GL) [4], measures of carbohydrate quality and quantity, respectively. Previous studies suggest that depression may arise from a decrease in brain serotonin neurotransmission [5,6,7,8,9]. Thus, high dietary GI or GL might be expected to alleviate depressive symptoms [6, 7] through the stimulation of insulin secretion, which facilitates the transport of tryptophan—a precursor of serotonin [7]—in the brain, and the consequent increase in serotonin synthesis [8, 9]. Conversely, there is some evidence to show that insulin resistance, potentially caused by a high GI and GL diet, may be associated with a pattern of volumetric and neurocognitive deficits in diabetic populations that are highly similar to that reported in populations of individuals with major depressive disorder [10]. In addition, high dietary GI and GL can cause rapid late postingestive decrease in blood glucose with a tendency to hypoglycemia, eliciting central dysfunction and depression, which might contribute to the higher prevalence of depression among patients with diabetes [11].

Epidemiologic evidence on the association between dietary GI and GL and depressive symptoms is not only limited, particularly in Asian populations, but also inconclusive [12,13,14,15,16,17]. The Japanese dietary pattern has several characteristics seldom observed in Western populations, including high intakes of rice, soybean products, fish, seaweeds, and green tea [18], in addition to high dietary GI and GL values [19,20,21]. Median values of dietary GI and GL in Japanese people (ranges of median values 65–67 and 141–185, respectively; GI for glucose = 100) [19,20,21] are consistently higher than those observed in Western countries (ranges of median values 50–61 and 80–140, respectively) [22], mainly because white rice, a food with a high GI, is the major contributor to dietary GI and GL in Japan, at 46–64% [19, 20]. Thus, the association between dietary GI and GL and depressive symptoms may differ between Japanese and Western populations. To our knowledge, however, this possible association has not been studied in Japanese.

Here, we conducted a cross-sectional study to examine the association between dietary GI and GL and depressive symptoms in a group of young and middle-aged Japanese women with relatively high dietary GI and GL [23].

Subjects and methods

Survey design and analytic sample

The present cross-sectional study was based on self-reported information obtained from students and their mothers in the Three-generation Study of Women on Diets and Health—a questionnaire survey. A detail of the study design and survey procedure has been published elsewhere [23, 24]. Briefly, a total of 7016 dietetic students from 85 higher education institutions in 35 of 47 prefectures in Japan were asked to answer two questionnaires on dietary habits and lifestyle factors during the orientation session or first lecture designed for freshmen in April 2011 or 2012. Each student was also requested to directly distribute the questionnaires to his or her mother (as well as grandmother) and to invite them to join the study. In total, 4933 students (4654 women and 277 men; response rate: 70.3%) and 4044 mothers (response rate: 57.6%) answered both questionnaires. The protocol of the study was approved by the Ethics Committee of the University of Tokyo Faculty of Medicine. Written informed consent was obtained from all participants, and also from a parent for participants aged <20 years.

In the present study, we selected female students aged 18 years (n = 4065) and mothers aged ≤65 years (n = 4034) for the analyses of young and middle-aged women, respectively. We then excluded students who answered the questionnaires after 19 May to minimize the influence of dietetic education (n = 56); subjects living in eastern Japan who participated in the 2011 data collection period, because it was assumed that they could not report their usual dietary habits and lifestyle due to the occurrence of the Great East Japan Earthquake in March 2011 (n = 39 and 63 for young and middle-aged women, respectively); and those from one institution at which the response rate was extremely low (2%; n = 2 for both young and middle-aged women). We further excluded those with missing information on the variables of interest (n = 5 and 143 for young and middle-aged women, respectively). Consequently, the final sample sizes were 3963 and 3826 for young and middle-aged women, respectively.

Dietary assessment

Dietary habits during the preceding month were assessed using a comprehensive diet history questionnaire (DHQ) [25,26,27]. Responses to the DHQ (and to the lifestyle questionnaire) were checked by survey staff at the study center. If any missing or implausible responses were given to questions considered essential for analysis, the subject was asked to complete the questions again. Details of the structure and calculation method for dietary intake of the DHQ have been published elsewhere [25,26,27]. Briefly, the DHQ is a structured, self-administered questionnaire that asks about the consumption frequency and portion size of selected foods commonly consumed in Japan, as well as general dietary behavior and usual cooking methods [27, 28]. Estimates of the daily intake of foods (151 items), energy, and selected nutrients were calculated using an ad hoc computer algorithm for the DHQ, which was based on the Standard Tables of Food Composition in Japan [29]. To minimize the influence of dietary misreporting [30, 31], energy-adjusted values of dietary variables were calculated based on the residual method [32]. The validity of the DHQ has previously been investigated in 92 women aged 31–69 years using 16-day weighed dietary record as Ref. [26]. With regard to the present analysis, the Pearson’s correlation coefficients between the two methods were 0.58 for EPA, 0.53 for DHA, and 0.55 for folate [26].

Calculation of dietary glycemic index and glycemic load

Calculation methods for dietary GI and GL [19, 20] as well as the GI value of each food item [23] in the DHQ have been published previously. Briefly, a total of 62 (of 151) food items were directly matched to foods available in the GI database, which was developed based on previous publications [33,34,35,36,37,38,39]. When more than one GI value was available, the mean value was used. Glucose was used as the reference (GI for glucose = 100). Food items for which a GI value had not been determined in previous publications (n = 10) were assigned the GI value of the closest comparable food (based mainly on macronutrient and fiber content). We assigned a GI value of 0 to other food items, including alcoholic beverages (due to no established GI values for them; n = 6) [40,41,42] and foods with low or no available carbohydrate content (<3.5 g per standard portion size; n = 73). Dietary GL was calculated as the sum of the product of the GI value of each food item and the amount of available carbohydrate intake from that food item (g/day) divided by 100. Dietary GI was calculated as dietary GL divided by the total amount of available carbohydrate intake (g/day) and multiplied by 100. For analysis, values of dietary GI and GL were energy-adjusted by the residual method [32]. In a group of 92 Japanese women aged 31–69 years, Pearson’s correlation coefficient between the DHQ and 16-day weighed dietary record was 0.53 for dietary GI and 0.58 for dietary GL [20] suggesting satisfactory validity of the DHQ in terms of dietary GI and GL.

Assessment of depressive symptoms

Depressive symptoms were assessed using a Japanese version [43] of the Center for Epidemiologic Studies Depression (CES-D) scale [44]. This scale consists of 20 questions that address 6 symptoms of depression, including depressed mood, guilt or worthlessness, helplessness or hopelessness, psychomotor retardation, loss of appetite, and sleep disturbance experienced during the preceding week. Each question was scored from 0 to 3 according to the frequency of the symptoms, and the total CES-D score ranged from 0 to 60. The criterion validity of the CES-D scale has been well-established in Western [44] and Japanese [43] subjects. Depressive symptoms were defined as present when a subject had a CES-D score ≥16 [43, 44].

Assessment of other variables

Age at the time of the survey was calculated based on birth date. Body mass index (BMI) was calculated as body weight (kg) divided by the square of body height (m). Information on current smoking (yes or no), current alcohol drinking (yes or no), medication use (yes or no) and level of stress (very low, low, normal, high, or very high) was also used. Physical activity was computed as the total metabolic equivalent-hour score per day on the basis of the frequency and duration of seven activities (walking, bicycling, standing, running, high-intensity activities, sleeping, and sedentary activity) over the preceding month [45]. Living alone (yes or no) was considered for young women (but not for middle-aged women because almost all lived with family). Education level (low (≤12 years), middle (13–15 years), and high (≥16 years)) and current marital status (yes or no) were considered for middle-aged women only. Dietary reporting status was evaluated on the basis of the ratio of reported energy intake (EI) to basal metabolic rate (BMR) (Goldberg’s cut-off) [46]. A detailed description of the procedure has been published elsewhere [23]. Briefly, BMR was calculated according to an equation specifically developed for Japanese women [23]. Subjects were identified as plausible, under-, and over-reporters of EI according to whether the individual’s ratio was within, below, or above the 95% confidence limits for agreement between EI:BMR and a physical activity level for sedentary lifestyle (i.e., 1.55) [46]. Consequently, under-, plausible, and over-reporters were defined as having an EI:BMR <1.09, 1.09–2.21, and >2.21, respectively.

Statistical analysis

All statistical analyses were performed for young and middle-aged women separately using the SAS statistical software, version 9.4 (SAS Institute Inc., Cary, NC, USA). All reported P values are two-tailed, and P < 0.05 was considered statistically significant. Descriptive data are presented as means and standard deviations (SDs) for continuous variables and percentages of subjects for categorical variables. Differences in characteristics between subjects with and without depressive symptoms were evaluated using an independent t test (for continuous variables) and Chi-square test (for continuous variables). Dietary GI and GL energy-adjusted by the residual method were categorized at quintile points on the basis of the distribution of young and middle-aged women. Associations between selected variables and dietary GI and GL (in quintiles) were examined by a linear trend test (for continuous variables) and a Mantel–Haenszel Chi-square test (for categorical variables). Crude and multivariate-adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for depression symptoms for each quintile category of dietary GI and GL were calculated using logistic regression analysis. The lowest quintile category of dietary GI and GL was used as a reference category. Potential confounding factors considered were BMI, current smoking, alcohol drinking, medication use, self-reported level of stress, dietary reporting status, and physical activity. For young women, additional adjustment was made for living alone. For middle-aged women, additional adjustment was made for age, education, and marital status. Further adjustment was made for intakes of energy, EPA + DHA, and folate. We tested for linear trends with increasing levels of dietary GI and GL by assigning each subject the median value for the category and modeling this value as a continuous variable. Associations between intakes of selected food groups, namely white rice; sugar and confectioneries; bread; noodles; soft drinks; fruit and vegetable juice; dairy products; and fruit, and depressive symptoms were also examined. These food groups were selected because they were major contributors to the inter-individual variation in dietary GI or GL in the present population [23]. All the analyses were repeated after excluding under- and over-reporters of EI.

Results

Basic characteristics of subjects are shown in Table 1. All of the young women were 18 years old while the mean (SD) age of the middle-aged women was 47.8 (4.1) years (age range 34–64 years). The prevalence of depressive symptoms was 50.2% for young women and 27.3% for middle-aged women. The mean (SD) values of energy-adjusted dietary GI and GL were, respectively, 64.9 (4.3) and 142.0 (27.4) for young women and 65.0 (4.1) and 142.2 (29.5) for middle-aged women. White rice contributed about half of dietary GL in both age groups, followed by sugar and confectioneries, which contributed about one-fifth of dietary GL. Pearson’s correlation coefficients between dietary GI and GL were 0.54 for young women and 0.52 for middle-aged women. Both dietary GI and GL were associated positively with carbohydrate intake (Pearson’s correlation coefficients: 0.20 to 0.92) but inversely with intakes of protein (−0.23 to −0.64), fat (−0.23 to −0.86), and dietary fiber (−0.19 to −0.33) in both age groups. Young women with depressive symptoms were more likely to use medication, have a higher level of stress, and under- or over-report EI, and had lower means of folate intake and dietary GI. Middle-aged women with depressive symptoms were more likely to be younger, currently smoke, use medication, have lower education, be unmarried, have a higher level of stress, over-report EI, and be physically inactive, and had a lower mean of folate intake. Depressive symptoms were also associated with both lower and higher EI in both age groups.

Table 1 Basic characteristics of young and middle-aged Japanese women
Full size table

Selected characteristics of young women according to quintile of dietary GI and GL are shown in Table 2. Subjects in the higher quintiles of dietary GI and GL were more likely to live alone and under-report EI and less likely to drink alcohol. Those in the higher quintiles of GI were also less likely to use medication. Dietary GI and GL were inversely associated with physical activity and intakes of EPA + DHA and folate. There was also a positive association between dietary GI and BMI.

Table 2 Selected characteristics of young Japanese women according to quintile (Q) of energy-adjusted dietary glycemic index (GI) and glycemic load (GL) (n = 3963)
Full size table

Selected characteristics of middle-aged women according to quintile of dietary GI and GL are shown in Table 3. Subjects in the higher quintiles of dietary GI and GL were more likely to have lower education and to under-report EI and had a lower mean folate intake. Subjects in the higher quintiles of dietary GI were more likely to be younger and less likely to use medication. Subjects in the higher quintiles of dietary GL were more likely to be older and have a higher level of stress and less likely to smoke and drink alcohol and had a lower mean intake of EPA + DHA.

Table 3 Selected characteristics of middle-aged Japanese women according to quintile (Q) of energy-adjusted dietary glycemic index (GI) and glycemic load (GL) (n = 3826)
Full size table

Associations between dietary GI and GL and depressive symptoms are shown in Table 4. In young women, an inverse association between dietary GI and depressive symptoms was observed in all models. This inverse association became stronger after adjustment for not only non-dietary, but also dietary confounding factors (model 3). The multivariate-adjusted ORs (95% CI) in the lowest to highest quintiles of dietary GI were 1 (reference), 0.71 (0.57, 0.88), 0.94 (0.76, 1.18), 0.75 (0.60, 0.93), and 0.66 (0.52, 0.82), respectively (P for trend = 0.001). In middle-aged women, an inverse association between dietary GI and depressive symptoms was observed only after adjustment for not only non-dietary, but also dietary confounding factors (model 3). The multivariate-adjusted ORs in the lowest to highest quintiles of dietary GI were 1 (reference), 0.81 (0.64, 1.01), 0.88 (0.70, 1.10), 0.85 (0.68, 1.07), and 0.75 (0.60, 0.96), respectively (P for trend = 0.046). Conversely, there was no association between dietary GL and depressive symptoms in either age group. In young women (Supplemental Table 1), higher intakes of sugar and confectioneries, soft drinks, and fruit and vegetable juice and lower intakes of dairy products were associated with a higher prevalence of depressive symptoms, with no associations for white rice, bread, noodle, and fruit. In middle-aged women (Supplemental Table 2), only intakes of soft drinks and noodles showed positive associations with depressive symptoms. Similar findings with regard to the associations between dietary GI and GL and depressive symptoms were observed after excluding under- and over-reporters of EI (n = 1001 in young women and 725 in middle-aged women; Supplemental Table 3), although the inverse association for GI did not reach statistical significance in middle-aged women.

Table 4 Odds ratios (95% confidence intervals) for depressive symptoms according to quintile (Q) of energy-adjusted dietary glycemic index (GI) and glycemic load (GL) in young and middle-aged Japanese women
Full size table

Discussion

In this cross-sectional study in young and middle-aged Japanese women, in whom dietary GI and GL are not only relatively high, but also primarily determined on the basis of the GI of white rice [23], higher dietary GI was associated with a lower prevalence of depressive symptoms, whereas no association was seen for dietary GL. To our knowledge, this is the first study to examine dietary GI and GL in relation to depressive symptoms in Asian populations.

The number of epidemiological studies of the associations between dietary GI and GL and depressive symptoms is limited, and the results have been inconsistent [12,13,14,15,16,17]. A prospective study in 69,654 postmenopausal US women aged 50–79 years showed that dietary GI (but not GL) was positively associated with the incidence of depression after adjustment for age, race, education, annual income, BMI, diabetes, hypertension, myocardial infarction, stroke, cardiovascular disease, cancer, Alzheimer disease, hormone replacement therapy, physical activity, alcohol intake, smoking status, stressful life events, social support, energy-adjusted intakes of SFAs, MUFAs, PUFAs, trans fat, dietary fiber, fruit, vegetables, legumes, and nuts/seeds, and Healthy Eating Index score [12]. A cross-sectional study from Australia similarly showed a positive association between dietary GI (but not GL) and depressive symptoms assessed by the Mental Health Index in 1981 adults (≥55 years of age) after adjustment for age, sex, cognitive impairment, walking disability, receiving pension, antidepressant use, and previous history of stroke and arthritis [13]. In the same study, however, no association was observed in 1528 adults (≥60 years of age) when depressive symptoms were assessed by the CES-D-10 [13]. In addition, positive associations between dietary GI and GL and depression were observed in a cross-sectional study of 976 homebound elderly US subjects (mean age 75.3 years) after adjustment for age, race, sex, education, BMI, and diabetes [14]. In a cross-sectional study in 3363 Iranian adults (mean age 36 years), however, dietary GI was positively associated but dietary GL was inversely associated with depression after adjustment for age, sex, marital status, education, physical activity, smoking, intakes of energy, ω-3 fatty acid, dietary fiber, magnesium, cobalamin, total folate, and pyridoxine, and BMI [15]. Additionally, dietary GL (but not GI) was inversely associated with depression in 140 elderly Spanish people aged 65–90 years after adjustment for sex, Barthel index, and Cambridge Cognitive Examination, as well as protein intake (GI only) [16]. No association was seen between dietary GI and GL during pregnancy and postpartum depression in 865 Japanese women [17]. The inverse association for dietary GI but no association for GL which we observed are somewhat different from these previous findings. Dietary GI in our population (mean 64.9 for young and 65.0 for middle-aged women; median 65.4 for young and 65.2 for middle-aged women), which is very similar to those observed in previous Japanese studies (range of median 65–67) [19,20,21], was on average higher than that observed in these non-Asian studies (mean 50–53 [16]; median 51.7 [12], 55.6 [14], and 62.2 [15], shown depending on the values available). Additionally, attention should be paid to differences in the major food contributors to dietary GI and GL. In Western diets these are based on a variety of foods, including bread, pasta, potatoes, sweetened beverages, sugar, fruit, and breakfast cereals [47,48,49]. In contrast, consistent with previous studies [19,20,21], the major contributor in Japanese diets was white rice, which accounted for about 50% of dietary GI and GL in the present population [23]. As a result, dietary GI was associated positively with only white rice and inversely with all other foods, including sugar and confectioneries, noodles, and fruit and vegetable juice [23]. In the present analysis, while white rice showed no association with depressive symptoms, higher intakes of soft drinks, sugar and confectioneries (only young women), fruit and vegetable juice (only young women), and noodles (only middle-aged women) and lower intakes of dairy products (only young women) were associated with a higher prevalence of depressive symptoms. These associations might contribute to the inverse association between dietary GI and depressive symptoms observed here. In any case, the differences in the distribution of and the food contributors to dietary GI and GL may at least partly explain the discrepant findings, in addition to differences in the characteristics and lifestyles of the populations examined, methods for assessing depression used, and potential confounding factors considered.

In this study, we found an inverse association with depressive symptoms for dietary GI only, and not GL. At least in the present population, carbohydrate quality only might be important, rather than both quality and quantity. Alternatively, the effect of dietary GI may be at least partly due to the overall dietary patterns associated with GI [50,51,52], given that dietary GI, unlike GL, appears to reflect more dimensions of diet than just carbohydrates, such as the combination of foods consumed [52]. Nonetheless, because previous studies observed that dietary GL was strongly associated with carbohydrate and dietary fat [50, 51, 53, 54], any effects of dietary GL cannot be separated from those of the macronutrient composition and overall diet quality in this observational study.

There are several limitations of our study. First, the cross-sectional design did not permit the assessment of causality owing to the uncertain temporality of the assessment. As a result, we cannot exclude the possibility that depressive symptoms may lead to a lower consumption of low-GI foods because of a loss of appetite or lead to a higher consumption of some foods such as sugar and confectioneries (which decrease dietary GI in the present population [23]) because of an increased appetite, for example. Actually, depressive symptoms were associated with both lower and higher EI in this study. The relationship between dietary intake and mental health is complex and likely bidirectional, and the temporal direction of the association between diet quality and depressive symptoms could be both ways [55]. In any case, only a prospective study would provide better understanding of the relationship between dietary GI and GL and depressive symptoms.

Second, given the proportion of current Japanese adolescents who study in college or university (57%) [56], our participants (i.e., dietetic students and their mothers) are likely to have a relatively high socio-economic status. Further, dietetic students may be more conscious of their diet than the general population, although the present survey was carried out in most institutions within 1 month after the dietetic course began to minimize the influence of dietetic education. Thus, our results might not be applicable to the general Japanese population.

Third, all self-reported dietary assessment methods are subject to both random and systematic measurement errors. To minimize these, we assessed dietary habits during the preceding month using a well-established dietary assessment questionnaire with reasonable validity in terms of commonly studied nutritional factors (DHQ) [25,26,27], and also used energy-adjusted dietary variables [31]. Further, our DHQ is not designed specially to measure dietary GI and GL, as is the case of other dietary assessment questionnaires [12,13,14,15, 48, 51,52,53]. Nonetheless, the satisfactory validity of the DHQ regarding dietary GI and GL against a 16-day weighed dietary record as described above may provide some assurance on this point [20]. Additionally, as mentioned above, exclusion of EI misreporters did not change the results materially, which may support the robustness of the present findings.

Forth, depressive symptoms were assessed using a validated questionnaire (i.e., CES-D) [43, 44] rather than structured diagnostic interviews. The absence of a clinical diagnosis may have led to the inclusion of subjects with chronic fatigue syndrome or atypical depression. Nevertheless, a similar prevalence of depressive symptoms has also been observed among university students (50%) [57] and in a nationally representative sample of middle-aged women (30–59 years of age; 31%) [58] using the same methodology (CES-D score ≥16). A similar or somewhat lower prevalence (using the same methodology) has been reported from Western countries, such as the US (24% [59] and 38% [60]) and France (33% [61]). Evidence suggests an over-rating of symptoms by Japanese teenagers compared with American counterparts [62], which may explain an excessively high prevalence of depressive symptoms in young Japanese women. Nevertheless, we note that the use of higher cut-off values (i.e., CES-D score ≥19 or 23) in the analysis of young women did not change the associations between dietary GI and GL and depressive symptoms (data not shown). Finally, although adjustment for a variety of potential confounding variables was done, residual confounding could not be ruled out.

In conclusion, this cross-sectional study in young and middle-aged Japanese women, in whom dietary GI and GL are not only relatively high, but also determined on the basis of a high GI food white rice [23], revealed an inverse association between dietary GI and depressive symptoms but not for dietary GL. Further research, preferably with a prospective design, is needed to confirm our findings in other Japanese and Asian populations whose staple food is white rice.