Sugar intake from sweet food and beverages, common mental disorder and depression: prospective findings from the Whitehall II study

Intake of sweet food, beverages and added sugars has been linked with depressive symptoms in several populations. Aim of this study was to investigate systematically cross-sectional and prospective associations between sweet food/beverage intake, common mental disorder (CMD) and depression and to examine the role of reverse causation (influence of mood on intake) as potential explanation for the observed linkage. We analysed repeated measures (23,245 person-observations) from the Whitehall II study using random effects regression. Diet was assessed using food frequency questionnaires, mood using validated questionnaires. Cross-sectional analyses showed positive associations. In prospective analyses, men in the highest tertile of sugar intake from sweet food/beverages had a 23% increased odds of incident CMD after 5 years (95% CI: 1.02, 1.48) independent of health behaviours, socio-demographic and diet-related factors, adiposity and other diseases. The odds of recurrent depression were increased in the highest tertile for both sexes, but not statistically significant when diet-related factors were included in the model (OR 1.47; 95% CI: 0.98, 2.22). Neither CMD nor depression predicted intake changes. Our research confirms an adverse effect of sugar intake from sweet food/beverage on long-term psychological health and suggests that lower intake of sugar may be associated with better psychological health.

associated with increased circulating inflammatory markers, which may depress mood 20,21 . Thirdly, high sugar diets could induce hypoglycaemia through an exaggerated insulin response and thereby influence hormone levels and potentially mood states 22 . Fourthly, addiction-like effects of sugar suggest dopaminergic neurotransmission mechanisms might connect frequent sugar intake with depression [23][24][25] . Lastly, obesity could be a mediating factor between a sugar-dense diet and depression 26,27 not only via inflammatory but also psychosocial factors like weight discrimination 28 .
The aim of this study is to investigate whether sugar intake from sweet food/beverages is positively associated with the risk of both incident and recurrent mood disorders, and to establish the role of the reverse effect in the Whitehall II cohort, using prospective, repeat measures data collected over a 22 year period.

Methods
Study cohort. The Whitehall Study II consists of non-industrial civil servants, who were recruited in London at age 35 to 55 years during 1985-1988 (phase 1). The initial sample size was 10,308 individuals (33.1% female and 66.9% male). The participants were followed up via questionnaire in [1989][1990]  . In phases 1, 3, 5, 7, 9 and 11 they were additionally invited for screening in a research clinic 29 . Phase 10 (2011) consisted of a smaller sample of participants used for a pilot study. The study was approved by the Joint UCL/UCLH Committee on the Ethics of Human Research and carried out in accordance with the ethical principles set out in the Declaration of Helsinki. Further all participants have been asked for informed consent at every follow-up.
Ascertainment of sugar intake from sweet food/beverages. Diet was assessed at phases 3, 5, 7 and 9 using a 127-item machine-readable semi-quantitative food frequency questionnaire (FFQ) which originates from the tool used in the US Nurses' Health Study, a self-administered questionnaire on habitual diet over the past 12 months 30,31 . In order to reflect most diets in the UK it has been modified and anglicized 32 . This FFQ has been validated against a 7 day diet diary in a stratified random sample of 865 participants in the Whitehall Study II at collection phase 3 30 . Sweet food and beverage intake was measured with 15 items such as cakes, biscuits, added sugar to coffee or tea, and fizzy soft drinks (see Supplementary Table S1). Sugar intake was calculated by multiplying sweet food/beverage consumption frequencies per day by their sugar content and portion size based on McCance and Widdowson's The Composition of Foods, 5th edition 33 . Depressive symptom assessment. The 30-item General Health Questionnaire (GHQ) measures depressive and somatic symptoms over the past two weeks 34 . Caseness was defined as reporting ≥5 symptoms and is referred to as common mental disorder (CMD). This measure was included in follow-up questionnaires at all phases apart from phase 4. In addition, the 20-item Center of Epidemiologic Studies Depression Scale (CES-D), a self-report measure of depressive symptoms in the general population over the past week 35 , was administered at phases 7, 9 and 11. Individuals scoring ≥16 were considered cases of depression 36 . Lastly, a clinical interview using the Revised Clinical Interview Schedule (CIS-R) was administered at phase 11 with participants assessed according to International Classification of Diseases (ICD-10) F32 criteria. The computerized self-completion version of the CIS-R included questions on depressive symptoms that were present for at least 2 weeks [37][38][39] . The GHQ and CES-D have been validated against the CIS-R in this cohort and showed high sensitivity and specificity in measuring depressive episodes 39 . Covariates. Potential confounders were chosen based on review of the literature and restricted to variables available at all phases used in the analyses. All estimates were initially adjusted for age, ethnicity (White/ South Asian/ Black) and sex, with an interaction of sex and age where both sexes included. Socio-demographic variables consisted of marital status (married/cohabiting, single or divorced/widowed) and last employment grade level within the civil service, (high, intermediate, low). Health behaviours included smoking (never, former, current), alcohol intake (none: ≤1 unit/weeks, moderate, heavy: ≥14 units/week) self-reported physical activity (vigorous, moderate and non/mild) 40 and duration of sleep (5 categories from ≤5 hours to ≥9 hours/day). Diet-related factors comprised energy intake, diet quality, fish, coffee and tea intake based on FFQ data. Energy intake was used to ascertain dietary misreporting. Misreporting was considered where the log ratio of energy intake to estimated energy expenditure was outside of 3 SD of the log mean. This definition was adopted by Mosdol et al. 2007 and based on basal metabolic rate equations of the Department of Health [41][42][43] . Since sugar intake from sweet food/ beverages was strongly correlated with energy intake (r = 0.61, P < 0.001), energy intake was adjusted for with the partition method by using energy intake from other foods 44 . Diet quality was assessed using the Dietary Approaches to Stop Hypertension (DASH) diet score modified by excluding a measure for sweet drinks 45 . DASH diet score, coffee and tea intake were analysed as continuous variables, fish intake per day as quintiles and all dichotomized for descriptive analyses. Body mass index (BMI) (kg/m 2 ) and central obesity (in women waist circumference ≥88 cm and in men ≥102 cm) were both measured by trained staff 46 . Physical health was defined as diabetes and cardiovascular disease (coronary heart disease and stroke, CVD) based on self-reports which were validated using the study clinical examination, Hospital Episode Statistics data, and by contacting general practitioners for confirmation when no other external source existed. Cancer was based on cancer registration data 29 . Finally, doctor diagnosis of depression was based on self-report at phases 1 to 4 and on self-reported antidepressant intake at all phases after phase 4.

Statistical analysis.
At each phase, participants were included if they had answered at least 8 of the FFQ sweet food and beverage items 47 (less than 5% of eligible sample had one missing item and about 1% two or more), their ethnicity was known to be either White, Black or South Asian, and participants were not energy misreporters (see above). In addition, participants were also excluded from analyses if they had incomplete data on GHQ-, CES-D-or CIS-R caseness for outcome-specific analyses, respectively. Supplementary Fig. S1 shows how the included sample was reached (see Online).
Three binary outcomes were analysed, GHQ caseness, CES-D caseness and CIS-R caseness. Daily sweet food and beverage intake was modelled as sex-specific tertiles of sugar intake from sweet food/beverages based on the distribution at phase 3 (in men <39.5, ≥39.5 to <67.0 and ≥67.0 g/day; in women <30.0, ≥30.0 to <51.0 and ≥51.0 g/day). To describe the sample at phase 3, GHQ cases, non-cases and tertiles of sugar intake by covariate were compared. To examine the prospective association of sugar intake from sweet food and beverages, a random effects logistic regression model (REM) was performed using the STATA command xtlogit 48 , with exposures at phases 3, 5, 7 and 9 for GHQ caseness, and at phases 7 and 9 for CES-D caseness. The applicability of the REM was tested by introducing study phase-interactions and likelihood ratio tests (LRT). The prospective effect of sugar intake from sweet food/beverages on incident and recurrent CMD and depression was examined using REMs in 2, 5 and 10-year cycles 49 . Figure 1 shows the included phases for analyses using GHQ caseness as the outcome. For example, the association between sugar intake and GHQ status 2 years later was conducted by combining the associations between sugar intake at Phase 5 and incident GHQ caseness at Phase 6, and between sugar intake at Phase 7 and incident GHQ caseness at Phase 8. For all depression outcomes, incidence was assumed if no CMD was apparent at each baseline, and recurrence if CMD was apparent at each baseline. For the analyses of depression, two 5-year cycles (to Phase 9 and 11) and three 10 year cycles (to Phases 7, 9 and 11) were used. For clinical depression, one 5-year cycle and one 10-year cycle were used.
To check for reverse causation, that depressive symptoms may affect subsequent sugar intake from sweet food/ beverages, linear regression models of 5-year change and multinomial logistic regression for change groups were fitted for each cycle, from phases 3 to 5, 5 to 7 and 7 to 9, with CMD at phases 3, 5, 7 respectively, and for change from phase 7 to 9 with depression at phase 7. Normal distribution of change in sugar intake from sweet food/ beverages was verified using a histogram. Change groups were created by subtracting tertiles of sugar intake at baseline (t) from sugar intake from sweet food/beverages at follow-up (t + 5 y) and coding −2/−1 as decrease, 0 as no change and +1/+2 as increase in sugar intake from sweet food/beverages. All analyses were performed using Stata 14 50 . Interactions of CMD and depression with sex in the initial model (Model 0 per sex-specific tertile trend: adjusted for age and ethnicity) were tested using LRT since sex-differences have been reported in a prior study on the association of diet and depression in the Whitehall II cohort 51 . Further adjustments were grouped into four hierarchical models: baseline socio-demographic factors and health behaviours (Model 1), diet-related factors (Model 2), BMI and central obesity (Model 3), and physical health (Model 4). In sensitivity analyses, main analyses were repeated by: (a) excluding participants with unknown or reported doctor diagnosis of depression at each baseline (at phases 3/5/7/9: 166/156/193/209 individuals) and: (b) excluding participants with extreme values of sugar intake (>7 SD) at phases 3/5/7/9: 5/3/4/4 individuals. Table 1 shows the prevalence of CMD and tertiles of reported sugar consumption from sweet food/beverages according to covariates at phase 3. CMD was more prevalent in women: under 50-years old, divorced/widowed, physically inactive, current smokers and those with fewer hours of sleep. Women with CMD were more likely to be in a lower grade level in civil service (P < 0.001; not depicted). Sugar consumption was associated with socio-demographic factors, health behaviours, physical health and diet-related factors (Table 1). Unexpectedly, participants in the highest tertile of sweet food/beverage intake had the highest prevalence of normal weight and lowest prevalence of overweight and obesity as well as the lowest prevalence of abdominal obesity in men (both P = 0.002; not depicted).

Results
Incidence of CMD was around 9 to 15%, highest in the first cycle but did not differ greatly by cycle length. Depression and clinical depression incidence were approximately 8% and 2%, respectively. About 44% of participants who were CMD cases at baseline of each cycle remained recurrent CMD cases, 47% became recurrent depression cases and 58% recurrent clinical depression cases. Cross-sectional results. Cross-sectional analyses showed strong positive associations between sugar intake from sweet food/beverages and common mental disorder from the GHQ, as well as CES-D caseness, when adjusted for age, sex and ethnicity (Table 2). There was no evidence for any interaction with sex (P = 0.8 for GHQ and P = 0.7 for CES-D). The association with CMD was robust whereas for depression it was removed on adjustment for socio-demographic factors, health behaviours and diet-related factors ( In men, after adjustment for age and ethnicity, sugar intake was associated with incident CMD 2 and 5 years later (Table 3).
In further models, the association with 2-year incidence attenuated but the association with 5-year incidence remained ( Prospective analyses regarding the associations of sugar intake from sweet food/beverages and recurrent mood disorders showed no evidence for sex interaction for CMD, CES-D depression or clinical depression 5 years later. Sugar intake from sweet food/beverages was positively associated with recurrent depression after 5 years (Model 0, highest vs. lowest tertile OR: 1.81; 95% CI: 1.23, 2.66; P for tertile trend = 0.003, Table 4). The association was attenuated when adjusted for other diet-related factors. Moreover, there was some evidence that sugar intake from sweet food/beverages was associated with recurrent clinical depression in both sexes combined (highest vs. lowest tertile OR: 1.66; 95% CI: 0.96, 2.87 and P for tertile trend = 0.07) when adjusted for age, sex and ethnicity (Supplementary Table S3). This association attenuated when further factors were introduced to the model. No statistically significant association was found for sugar intake from sweet food/beverages and recurrent GHQ caseness after 2 and 5 years (Model 0 for CMD after 2 years, highest vs. lowest tertile OR: 1.05; 95% CI 0.76, 1.45; P for tertile trend = 0.83; for CMD after 5 years: 1.16; 95% CI 0.93, 1.46; P for tertile trend = 0.20).
Analyses of recurrent CMD, depression and recurrent clinical depression after 10 years showed no associations with sugar intake from sweet food/beverages. Sensitivity analyses excluding extreme values of sugar intake and excluding person-observations with self-reported doctor diagnosis at baseline attenuated the association of sugar intake from sweet food/beverages and recurrent depression slightly (before P for tertile trend 0.003 after 0.022 and 0.010, respectively). Similarly, associations with clinical depression weakened when participants with depression diagnosis at baseline were excluded (Model 0, Person observations = 573; cases = 78; P for tertile trend = 0.17).
Analysis of reverse causation. Sugar intake from sweet food/beverages decreased by 2.00 (SD 28.8; 95% CI 1.20, 2.79) grams per day from phase 3 to 5, by 3.44 (SD 28.0; 95% CI 2.59, 4.30) grams from phase 5 to 7 and by 1.57 (SD 26.0; 95% CI 0.91, 2.33) grams from phase 7 to 9, and was normally distributed. Mean 5-year change was approximately 31 g sugar from sweet food/beverages per day in the decrease group, −0.7 g in the stable intake group and 29 g in the increase group. Neither CMD, nor depression predicted 5-year changes in sugar intake (Table 5).

Discussion
The present long-term prospective study is the first to investigate the association of sugar consumption from sweet food/beverages with prevalent, incident and recurrent mood disorders, while also examining the effect  Table 3. Prospective association of sugar intake from sweet food/beverages and incident common mental disorder after 2 and 5 years in men a . Abbreviations: OR = Odds ratio, CI = Confidence interval, DASH = Dietary Approaches to Stop Hypertension. a Prospective association across phases 3, 5 for 2-year and 3, 5, 7, 9 for 5-year incident common mental disorder. b Common mental disorder measured using the 30-item General Health Questionnaire. c 2-year model 0 (655 events/5767 person observations): adjusted for age and ethnicity. d Model 1: additionally adjusted for marital status, last grade level in civil service, smoking, alcohol intake, physical activity, sleep duration. e Model 2: additionally adjusted for energy intake from other foods, modified DASH diet score, fish, coffee and tea intake. f 5-year model 0 (1448 events/12445 person observations): adjusted for age and ethnicity.
these disorders might have on subsequent habitual sugar intake. We found an adverse effect of higher sugar intake on mental health cross-sectionally and 5 years later in a study based on 23,245 repeated measures in men and women aged between 39 and 83. Further, we found an increased likelihood for incident CMD in men and some evidence of recurrent depression in both sexes with higher intakes of sugar from sweet food/beverages. These associations with incident CMD could not be explained by socio-demo graphic factors, other diet-related factors, adiposity and other diseases although the association with recurrent depression was explained by other diet-related factors. In our study we were able to exclude potential 'reverse causation' as the reason for the observed link between high sugar intake and low mood. Over years and decades, it could be that those susceptible to depression tend to increase their sugar intake. This group may tend to report higher consumption at study baseline even in the absence of depression at the time of the questionnaire, while having an increased risk of future depression compared to other participants 14,15 . However, there was no support for this alternative hypothesis, since the observed associations in our analysis were not the result of secondary changes in consumption of sugary food and drinks. Our study findings are consistent with the hypothesis that high sugar intake plays a causal role in the risks of both incident and recurrent depression and CMD.
Higher sugar intake from sweet food/beverages was associated with increased likelihood of incident CMD after 5 years in men. The association in men was in line with results from previous prospective studies in American and Spanish cohorts [9][10][11] . There are several potential explanations for the observed sex differences. First, the associations in men for incident CMD show a stronger effect with a bigger sample (comparing analysis of 2-year CMD with 5-year CMD), suggesting the lower number of female participants in our sample could have impaired the power of the analysis. Second, the results might reflect differences in pathways of depression by sex and type of depressive symptomatology [52][53][54] . Third, differences could be due to limitations of the study or to chance.
As described in the Introduction there are four potential mechanisms for an association of habitual sugar intake and subsequent depression risk. Sugar intake could increase depression risk over its potential influence on BDNF levels 16 and inflammation 20 which are both discussed as potential biological explanations for depression 17,21 . Furthermore postprandial hypoglycaemia 22 and addiction-like effects of sugar influencing neurotransmitters [23][24][25] could link sugar intake with low mood. The pathway of postprandial hypoglycaemia is also relevant in the context of Glycaemic index, which has been shown to be associated with depression prevalence and incidence 55,56 . However, it is a complex issue to tease apart the effects of a single nutrient in epidemiological studies since foods represent a mix of macro-and micronutrients. In this study associations were attenuated when adjusted for diet-related factors providing evidence of confounding and suggesting that the effect of sugar intake from sweet food/beverages could be partly explained by other components of the diet. Also, given that we analysed sugar intake from aggregated sweet foods and beverages we cannot rule out that certain types of foods and their particular components such as saturated fat content may have affected our findings. In our analysis, the association of sugar intake and recurrent depression was attenuated by measures of body fatness in participants without doctor diagnosis of depression at baseline supporting the hypothesis of an indirect effect mediated by adiposity 26-28 driving the association of sugar intake and recurrent depression.
Meanwhile, there are several sources of possible error. Our study was based on an occupational cohort but sugar intakes from sweet food/beverages were close to those reported previously in a representative cohort in the UK (approximately 40 grams), and Batty et al. showed that effects found in Whitehall II were comparable to those observed in population-representative cohorts 3,57 . A major limitation was the use of FFQ to derive diet data. FFQ data is subject to misreporting and underreporting, which have been found to differ by food group, depressive mood and BMI [58][59][60] . As reported previously in Whitehall, we showed a clear trend of lower sugar intake with higher BMI in men 60 . Nutrient content was based on food composition tables from 1991 and has to be considered as a source of error, since food composition especially of highly processed food is likely to change over the course of 18 years. In this long-term follow-up study, sugar intake from sweet food/beverages, which are consistently high in sugar content, has been used as the exposure measure. Compared to a measure of intake that includes processed foods 61 , this method may involve less information bias. Furthermore, this FFQ is meant to reflect habitual diet over the course of a year and therefore might not pick up short-term diet changes or occasional binge eating 30 . Although we adjusted for a number of potential confounders, we cannot rule out residual confounding through unknown or unmeasured factors. Finally, not all depression measures were obtained in all phases and selective dropout due to depressive symptoms might have influenced case numbers 62 .
In conclusion, our study provides evidence that sugar intake from sweet food/beverages increases the chance of incident mood disorders in men and limited evidence regarding recurrent mood disorders in both sexes. With a high prevalence of mood disorders, and sugar intake commonly two to three times the level recommended, our findings indicate that policies promoting the reduction of sugar intake could additionally support primary and secondary prevention of depression. To elucidate the association further, especially regarding observed sex differences our study should be replicated in representative prospective cohorts.

Common mental disorder b
At phase 3 -Sugar intake change: phase 3 to 5  Table 5. Association of common mental disorder and depression with subsequent 5-year change in sugar intake from sweet food/beverages. Abbreviations: No. = number, CI = confidence interval. a Change in sugar intake in cases compared with non-cases, adjusted for age, sex and ethnicity. b Common mental disorder measured using the 30-item General Health Questionnaire. c Depression measured using 20-item the Centre of Epidemiologic Studies Depression Scale.