Diet quality and mental health problems in adolescents ... - Springer Link

2 downloads 0 Views 201KB Size Report
Nov 18, 2012 - between the highest quintiles of both Healthy (OR 0.63,. 95 %CI 0.38–1.05) and Unhealthy (OR 1.75, 95 %CI. 1.00–3.06) diet scores and SDQ ...
Soc Psychiatry Psychiatr Epidemiol (2013) 48:1297–1306 DOI 10.1007/s00127-012-0623-5

ORIGINAL PAPER

Diet quality and mental health problems in adolescents from East London: a prospective study Felice N. Jacka • Catherine Rothon • Stephanie Taylor • Michael Berk • Stephen A. Stansfeld

Received: 21 February 2011 / Accepted: 2 November 2012 / Published online: 18 November 2012 Ó Springer-Verlag Berlin Heidelberg 2012

Abstract Purpose In this study, we aimed to examine the relationship between diet quality and depression in a prospective study of adolescents from varied ethnic and cultural backgrounds. Design In this prospective cohort study, data were collected at two time points (2001 and 2003) from nearly 3,000 adolescents, aged either 11–12 years or 13–14 years, participating in RELACHS, a study of ethnically diverse and socially deprived young people from East London in the UK. Diet quality was measured from dietary questionnaires, and mental health assessed using the Strengths and Difficulties Questionnaire (SDQ) and the Short Mood and Feelings Questionnaire (SMFQ). Results In cross-sectional analyses, we found evidence for an association between an unhealthy diet and mental health

problems. Compared to those in the lowest quintile of Unhealthy diet score, those in the highest quintile were more than twice as likely to be symptomatic on the SDQ (OR 2.10, 95 %CI 1.38–3.20) after taking all identified confounders into account. There was also some evidence for a crosssectional inverse association between a measure of healthy diet and mental health problems. A prospective relationship between the highest quintiles of both Healthy (OR 0.63, 95 %CI 0.38–1.05) and Unhealthy (OR 1.75, 95 %CI 1.00–3.06) diet scores and SDQ scores at follow-up was also evident, but was attenuated by final adjustments for confounders. Conclusion This study is concordant with previous observational studies in describing relationships between measures of diet quality and mental health problems in adolescents. Keywords Diet  Nutrition  Mental health  Depression  Adolescents

F. N. Jacka (&)  M. Berk Deakin University, School of Medicine, Geelong, Australia e-mail: [email protected]; [email protected] M. Berk e-mail: [email protected] C. Rothon  S. A. Stansfeld Wolfson Institute of Preventive Medicine, Centre for Psychiatry, Barts and The London School of Medicine and Dentistry, Queen Mary University of London, London, UK e-mail: [email protected] S. A. Stansfeld e-mail: [email protected] S. Taylor Institute of Health Sciences Education, Barts and the London School of Medicine and Dentistry, Queen Mary University of London, London, UK e-mail: [email protected]

Abbreviations SDQ Strengths and Difficulties Questionnaire SMFQ Short Mood and Feelings Questionnaire OR Odds ratio CI Confidence interval BMI Body Mass Index

Background Recent epidemiological data have pointed to a relationship between the quality of habitual diet and the common mental disorders, depression and anxiety. The published studies in adults have shown that diet quality is inversely related to the likelihood of clinically significant depressive

123

1298

and anxiety disorders and symptoms in women cross-sectionally [1] and to the risk for incident depression in adults over time [2–4]. We have recently investigated the association between diet quality and depression in more than 7,000 young Australian adolescents (10–14 years), from a diverse range of socio-economic and demographic backgrounds, using the Short Mood and Feelings Questionnaire (SMFQ) [5]. In this study, we found that adolescents scoring higher on a measure of Healthy diet were less likely to report symptomatic depression, while those with an increased consumption of processed and ‘junk’ foods were more likely to report depression. These associations demonstrated a dose–response pattern and remained robust after adjustment for a wide range of potential confounding factors including dieting behaviours, socio-economic status, family factors and other health behaviours. Similarly, another study in Australian adolescents has reported an inverse association between consumption of fruits and leafy green vegetables and both internalising and externalising behaviours, and a positive association between a ‘Western’ dietary pattern and increased behavioural problems [6]. Finally, we have recently documented that diet quality is associated with adolescent mental health over time [7]. In this prospective study of more than 3,000 Australian adolescents, predominantly from higher socioeconomic backgrounds, better diet quality at baseline predicted better mental health at follow-up even after adjustment for mental health at baseline. Moreover, improvements in diet quality were mirrored by improvements in mental health, while reductions in diet quality were associated with declining psychological functioning. Importantly, the reverse causality hypothesis, that the reported associations reflect poorer eating habits as a consequence of mental health problems, was not supported by the data. These are robust prospective data demonstrating that diet quality is an independent risk factor for the development of adolescent mental health problems. However, the available studies are limited in that they are confined to adolescents predominantly from the same ethnic and cultural backgrounds. In this study, we aimed to examine the relationship between measures of diet quality and adolescent mental health in the Research with East London Adolescents: Community Health Survey (RELACHS) [8, 9]. This study focuses its investigations on adolescents from ethnically diverse and socially deprived backgrounds, and comprises data from more than one time point. As such, the posited relationship can be examined prospectively, as well as cross-sectionally, in a sample of adolescents with dietary habits likely to be diverse. We hypothesised that unhealthy dietary patterns would be related to increased mental health problems, and healthy dietary patterns related to reduced

123

Soc Psychiatry Psychiatr Epidemiol (2013) 48:1297–1306

mental health problems, both cross-sectionally and prospectively.

Methods Study design and setting The RELACHS cohort study [8, 9] was designed to investigate relationships between ethnicity, measures of social disadvantage and psychological distress in adolescents, and assessed adolescents recruited from three Local Education Authority boroughs in East London (Hackney, Newham and Tower Hamlets) in 2001. These participants were followed up in 2003. Participants The baseline sample (2001) comprised pupils from years seven (aged 11–12 years) and nine (aged 13–14 years) attending a representative sample of state funded schools in the three catchment boroughs. The RELACHS study sample was selected using two-stage stratified random sampling. Of the 3,322 pupils eligible for the study, the overall response rate at baseline was 84 % (2,790 pupils). Of these, 2,093 (75 %) were assessed at time two in 2003. Data collection and ethics Information about the study was given to teachers, parents and pupils a week before the school visits. Parents could choose to opt out their child. Pupils who had not been opted out were invited to take part and asked for written consent. Pupils could withdraw from the study at any time, and did not have to answer questions they did not want to. A team of researchers administered the questionnaire in classrooms in one 40–50 min session. Pupils provided selfreported data on a self-completion questionnaire. Physical measurements were taken by trained researchers. Pupils were monitored to ensure that they were not distressed. Ethical approval was given by the ethics committees of East London and the City, as well as relevant local education authorities. Measurements: exposures Dietary data were collected at the baseline assessment only. Questions on healthy eating were taken from the Health and Behaviours of Teenagers Study (HABITS), and included consumption of fruit and vegetables [10] and regularity of eating breakfast. Breakfast consumption was assessed with the question ‘‘Before going to school, how often do you have breakfast at home or school breakfast

Soc Psychiatry Psychiatr Epidemiol (2013) 48:1297–1306

club?’’ with answers ranging from never (0) to every day (3). Questions on fruit and vegetable consumption comprised the following: ‘‘About how many lots of fruit do you usually eat in a day (how many ‘‘lots’’ means ‘‘how many portions’’ e.g. one apple/small bunch of grapes)’’; and ‘‘About how many lots of vegetables do you usually eat in a day?’’ Answers on these three questions were summed to give a ‘Healthy’ diet score. Consumption of unhealthy foods including fast foods, snacks and biscuits high in saturated fats and sugars was assessed with the question ‘‘About how often do you eat or drink the following? (for each): (a) Crisps or savoury snacks; (b) sweets, ghee sweets or chocolate; (c) biscuits; (d) fried food, chips, samosas or bhajis, or fried English breakfast; (e) fizzy drinks e.g. Coke’’, with answers ranging from never (0) to more than once a day (4). These answers were summed to give an ‘Unhealthy’ diet score. Measurements: outcomes Depressive symptoms were measured using the Short Moods and Feelings Questionnaire (SMFQ) [11]. This includes 13 statements about emotions and behaviour over the past 2 weeks. The scores for the items were summed to produce an overall magnitude of symptoms, with a score of eight or above indicating the presence of depression. In the original validation against the Diagnostic Interview Schedule for Children—Depressive Scale, this threshold yielded a positive predictive value of 80 % and a negative predictive value of 68 % [11]. Psychological distress was measured by the 25-item self-report version of the Strengths and Difficulties Questionnaire [12]. The SDQ comprises five sub-scales: hyperactivity, conduct problems, emotional symptoms, peer problems and prosocial behaviour (reverse scored). A total SDQ score ranging from 0 to 40 was generated by adding together the scores for all of the scales, apart from prosocial behaviour. The higher the total score, the higher the level of distress. A score of 18 was chosen as the threshold for a high scorer as this was equivalent to prevalence rates in national data using multi model assessments [13]. The measure has been used previously in ethnically mixed youth samples, which supports the SDQ as a valid instrument for ethnically diverse samples [13–15]. Measurements: confounders A number of confounding variables were tested for: gender, age, ethnicity, religion, length of time lived in UK, family structure, social deprivation (eligibility for free school meals, parental employment status, car ownership, over-crowding), general health, health behaviours (physical activity, alcohol use, smoking and drug use), dieting

1299

behaviour, body mass index (BMI) and family factors (parental conflict and social support). Questions on smoking and drinking (frequency) and drug use (ever used and how recently) were taken from the Office for National Statistics Survey for teenagers [16]. Questions on physical activity were taken from the Health Education Authority Survey [17]. Family social support came from the multidimensional scale of perceived social support [18]. Sociodemographic data (gender, religion, ethnicity and eligibility for free school meals) were drawn from the questionnaire and 2001 Census questions. Height and weight were measured according to a standardised protocol. Questions on dieting and perceived weight were also included. Data management All data management and analysis was carried out using Stata version 10.0. There were some missing data at baseline: 64 respondents did not have data on depression (2 %), 79 respondents did not have data on Healthy diet (3 %) and 134 respondents did not have data on Unhealthy diet (5 %). As the amount of missing data was relatively small for most of the key variables, it was considered reasonable to exclude pupils who did not have complete data. Statistical analyses Since the school was the primary sampling unit for the study, it was necessary to make adjustments for clustered survey design in the analyses. An equal number of classes were selected in each school regardless of school size; data were therefore re-weighted to adjust for unequal probability of selection. Cross-sectional analyses were carried out on the data collected in 2001. Descriptive statistics were generated, taking account of survey design. In the univariable analysis, crude odds ratios (ORs) were calculated for the association between each variable and psychological distress (SDQ) and depressive symptoms (SMFQ) separately using logistic regression. Confounding was assessed using classical (Mantel–Haenszel) and univariable logistic regression analyses. Effect modification was investigated by examining the results of the v2 test for homogeneity and by looking at the stratum specific ORs. Multivariable analysis was carried out using logistic/linear regression to examine the association between diet and mental health. Two outcomes were examined: psychological distress (SDQ) and depressive symptoms (SMFQ). Two main exposures were used in separate regression analyses: Healthy diet and Unhealthy diet. These two dietary scales were categorised into quintiles to reduce any impact of

123

1300

outliers, and to aid in identification of non-linear relationships. Potential confounders were identified on the basis of relationships between at least one exposure and one outcome and added to the models in groups (gender, dieting behaviour, health behaviours and family factors). If adding a confounder or group of confounders resulted in an improvement in model fit, as assessed by the Wald test, they were retained in the model. Multivariable longitudinal logistic regression analyses examined the impact of diet in 2001 on case level psychological distress and symptomatic depression in 2003. In secondary analyses, cases on the SDQ and SMFQ at baseline were excluded from the analyses, and continuous scores on the mental health questionnaires were used as outcome measures in linear regression analyses. Confounders were added to the model in the same way as for the cross-sectional analyses.

Results

Soc Psychiatry Psychiatr Epidemiol (2013) 48:1297–1306 Table 1 Characteristics of the sample at baseline N

Percentage (adjusted for survey design)

Male

1,356

48.8

Female

1,433

51.2

Year 7

1,381

49.0

Year 9

1,408

51.0

White

581

20.8

White other

161

5.9

Bangladeshi

690

25.6

Asian Indian

250

9.0

Pakistani

184

6.8

Black Caribbean

166

6.0

Black African

279

10.1

Black British

121

4.3

Mixed Ethnicity

193

7.0

Chin/Viet and other

124

4.5

325

11.5

Gender

Year group

Ethnicity

Religious group None Jewish

Survey

Christian Muslim/Islam

A total of 2,789 pupils completed the baseline survey in 2001. There was some evidence for lower response among pupils who were eligible for school meals (P = 0.055), and strong evidence for lower response amongst white pupils (P \ 0.0001). Seventy-five per cent of these, 2,093 pupils, were followed up in 2003. Some groups were less likely to be followed up: those depressed in 2001 (P = 0.002), girls (P = 0.005), those eligible for free school meals (P = 0.002) and white pupils (P B 0.0001). Table 1 describes the sample at baseline. Of the pupils, approximately 21 % were white and another 26 % Bangladeshi, with smaller proportions distributed amongst a multitude of other ethnicities. Almost 40 % of students reported both parents unemployed and nearly half were eligible for free school meals. At baseline, 9.2 % of boys and 11.3 % of girls were SDQ cases, while 10.2 % of boys and 13.2 % of girls were cases on the SDQ at follow-up. Of these, 8.4 % of boys and 10.1 % of girls were ‘new’ cases at follow-up. At baseline, 18.9 % of boys and 29.8 % of girls were cases on the SMFQ, while 19.6 % of boys and 34.6 % of girls were cases on the SMFQ at follow-up. Of these, 14.8 % of boys and 25.4 % of girls were new cases at follow-up. Healthy diet scores were normally distributed, while Unhealthy diet scores showed a negatively skewed distribution. Healthy diet scores ranged from 0 to 13 (SD ± 2.82), while Unhealthy ranged from 1 to 20 (IQR = 6).

123

Hindu Sikh Agnostic/Atheist/Don’t know Other Case on SDQ Not a case on SDQ

3

0.1

994

35.7

1,169

42.7

101

3.7

72

2.7

59

2.1

43

1.6

286

10.3

2,458

89.7

671

24.5

2054

75.6

0

466

17.2

1

274

10.0

2

782

29.0

3

636

23.2

4

552

20.6

0

498

18.6

1

476

18.1

2

492

18.4

3

506

19.1

4

683

25.8

Neither

1,004

37.5

At least one

1,681

62.5

Eligible

1,453

48.1

Not eligible

1,550

51.9

Case on SMFQ Not a case on SMFQ Healthy diet score

Unhealthy diet score

Either parent employed

Eligible for free school meals

Cross-sectional univariable analysis Table 2 reports the results of univariable analyses. Healthy diet scores were inversely related to both the SMFQ and

Soc Psychiatry Psychiatr Epidemiol (2013) 48:1297–1306

1301

Table 2 Odds of psychological distress in 2001 by dietary and other variables: results of univariable logistic regression analyses SDQ

Eligible for free school meals

SMFQ

OR

95 % CI

N

OR

95 % CI

N

1.03

0.77–1.38

2,522

1.00

0.83–1.20

2,505

2,479

1

BMI Below 85th percentile

1

2,462

Overweight (85th–95th percentile)

0.89

0.58–1.34

1.04

0.72–1.49

Obese (over 95th percentile)

1.26

0.91–1.76

1.45

1.17–1.81

Not trying to change weight Trying to lose weight

1 1.88

1.37–2.57

1 2.30

1.86–2.85

Trying to gain weight

2.00

1.24–3.21

1.90

1.38–2.61

Dieting behaviour 2,596

2,592

Physical activity h/week None

1

About 0.5 h

0.81

0.59–1.11

2,716

1 0.94

0.71–1.25

2,705

About 1 h

0.68

0.41–1.13

0.90

0.67–1.21

About 2–3 h

0.55

0.36–0.85

0.67

0.49–0.91

About 4–6 h

0.61

0.30–1.25

0.58

0.37–0.90

More than 7 h

0.68

0.34–1.34

0.48

0.30–0.77

Cigarette smoking Never smoked

1

Less than one per week

1.68

1.31–2.12

2,710

1 1.48

1.14–1.92

2,704

One or more per week

2.91

1.87–4.52

1.92

1.23–2.98

Did not drink 1–5.5 units

1 2.29

1.09–4.80

1 1.06

0.59–1.88

More than 6 units

6.33

2.36–16.92

3.18

1.31–7.68

Units of alcohol consumed last week 2,678

2,679

Ever tried drugs No

1

Yes

2.39

2,738 1.72–3.33

1 1.49

2,719 1.21–1.83

Parents argued or fought No

1

Yes

1.40

1.05–1.88

Family social support (continuous)

0.74

0.68–0.81

2,,489

1

2,719

1.49

1.21–1.84

2,670

1.75

1.42–2.15

2,670

1

2,666

Family social support Low

1

Moderate

0.48

0.34–0.67

0.43

0.35–0.52

High

0.46

0.33–0.66

0.35

0.27–0.43

SDQ. Unhealthy diet scores were positively related to the SDQ, with a dose–response pattern observed, but the relationship of Unhealthy diet scores to the SMFQ; whilst in the same direction, was less strong and not statistically significant. Cigarette smoking, alcohol consumption and drug use were also associated with increased risk of caseness on the mental health scales, as were family conflict and low family support. Dieting behaviour (actively trying to lose or gain weight) was also associated with mental health problems, although BMI was related only to the SMFQ and only for those identified as obese (over 95th

2,666

percentile). Eligibility for free school meals, as a measure of social deprivation [19] was not associated with either mental health scale. Multivariable analysis: baseline No effect modification was observed for gender (P [ 0.05). Potential confounders in analyses with SDQ as an outcome variable were identified in several categories: dieting behaviours, health behaviours and family factors. For those examining the SMFQ as an outcome, gender was

123

1302

Soc Psychiatry Psychiatr Epidemiol (2013) 48:1297–1306

Table 3 Odds of psychological distress on the SDQ in 2001 per quintile of Healthy and Unhealthy diet score: results of multivariable logistic regression analyses Crude odds

?Dieting

OR

OR

95 %CI

?Health behaviours*

?Family factors**

95 %CI

OR

OR

95 %CI

95 %CI

Healthy diet score (Quintiles) Least healthy

1

1

1

1

1 2

0.84 0.64

0.50–1.43 0.40–1.03

0.88 0.69

0.51–1.51 0.42–1.13

0.96 0.78

0.54–1.70 0.48–1.24

0.97 0.83

0.54–1.72 0.53–1.34

3

0.59

0.37–0.93

0.63

0.39–1.00

0.75

0.46–1.23

0.82

0.51–1.34

0.66

0.47–0.94

0.70

0.49–0.99

0.87

0.61–1.25

1.00

0.69–1.44

Most healthy

\0.001

Wald p value (N = 2,383)

\0.001

0.003

Unhealthy diet score (Quintiles) Most healthy

1

1

1

1

1

1.56

0.80–3.01

1.57

0.80–3.09

1.54

0.80–2.96

1.56

0.82–2.97

2

1.66

1.05–2.62

1.75

1.10–2.76

1.60

0.97–2.64

1.60

1.00–2.57

3

2.14

1.24–3.68

2.34

1.37–4.00

2.10

1.19–3.69

2.17

1.22–3.83

2.15

1.44–3.21

2.43

1.64–3.58

2.07

1.36–3.15

2.10

1.38–3.20

Least healthy Wald p value (N = 2,339)

\0.001

0.003

0.001

* Health behaviours: smoking, alcohol consumption, drug use and physical activity ** Family factors: parental conflict and family social support

also a confounder, but health behaviours were not. Tables 3 and 4 report the multivariable regression analyses for both the SDQ and the SMFQ at each stage of adjustments. There was strong evidence for an association between Unhealthy diet scores and caseness on the SDQ both before and after adjustments (Table 3). Compared to those in the lowest quintile for Unhealthy diet scores, those in the highest quintile were more than twice as likely to be a case on the SDQ, both before and after adjustments. After adjustment for confounders, there was also evidence for an association between higher Unhealthy diet scores and increased odds of symptomatic depression on the SMFQ. This relationship was only apparent after adjusting for dieting and family factors, suggesting that these were acting as suppressor variables in the analyses (Table 4). There was weaker evidence for an association between healthy diet scores and the mental health outcomes. The inverse association between healthy diet scores and the SDQ observed in univariable analyses was fully explained by adjustments for other health behaviours (Table 3), while the relationship of healthy diet scores to the SMFQ was somewhat attenuated by adjustments for gender, then fully attenuated by adjustments for family factors (Table 4). Multivariable analysis: prospective In analyses examining Unhealthy diet scores at baseline as predictors of case level mental health outcomes at follow-

123

up, a relationship was observed with the SDQ for those in the highest quintile of unhealthy diet, but not the SMFQ. Prior to adjustments, those in the highest quintile of Unhealthy diet scores at baseline were significantly more likely to be a case on the SDQ at follow-up (OR 1.75, 95 % CI 1.00–3.06). However, after all adjustments, including adjustment for caseness at baseline, the relationship was in the same direction but no longer significant (OR 1.50, 95 % CI 0.80–2.81). Weak inverse relationships between Healthy diet scores and mental health at follow-up were observed for the SMFQ [Q1 = reference (Q2: OR 0.82, 95 % CI 0.60–1.11; Q3: OR 0.65, 95 % CI 0.46–0.92; Q4: OR 0.70, 95 % CI 0.52–0.94; Q5: OR 0.75, 95 % CI 0.55–1.04)], but were fully attenuated by adjustments. A similar pattern of inverse associations was observed between Healthy diet scores and the SDQ at follow-up [Q1 = reference (Q2: OR 0.74, 95 % CI 0.39–1.41; Q3: OR 0.60, 95 % CI 0.33–1.11; Q4: OR 0.61, 95 % CI 0.39–0.97; Q5: 0.63, 95 % CI 0.38–1.05)] in univariable analyses, but these were also attenuated by adjustments for other factors. Results of analyses with continuous scores on the SDQ as the outcome variable revealed that diet scores at baseline were inversely related to SDQ scores at follow-up until final adjustments for family factors, wherein only the highest quintile of healthy diet score maintained a relationship of borderline significance with lower SDQ scores (ß = -0.91, 95 % CI -1.84–0.01). Once again, only the highest quintile of Unhealthy diet score demonstrated a

Soc Psychiatry Psychiatr Epidemiol (2013) 48:1297–1306

1303

Table 4 Odds of psychological distress on the SMFQ in 2001 per quintile of Healthy and Unhealthy diet score: results of multivariable logistic regression analyses Crude odds

?Gender

OR

OR

95 %CI

?Dieting 95 %CI

OR

?Family factors** 95 %CI

OR

95 %CI

Healthy diet score (Quintiles) Least healthy

1

1

1

1 2

0.72 0.64

0.50–1.04 0.48–0.85

0.77 0.68

0.54–1.08 0.51–0.91

0.80 0.73

0.56–1.12 0.54–1.00

0.81 0.83

0.58–1.14 0.61–1.14

3

0.60

0.46–0.79

0.64

0.50–0.83

0.68

0.51–0.91

0.80

0.60–1.07

0.70

0.50–0.99

0.78

0.56–1.09

0.82

0.57–1.18

1.05

0.71–1.55

Most healthy

\0.001

Wald p value (N = 2,382)

\0.001

\0.001

Unhealthy diet score (Quintiles) Most healthy

1

1

1

1

1

1.37

1.02–1.84

1.39

1.03–1.87

1.39

1.01–1.91

1.41

1.03–1.92

2

1.19

0.83–1.72

1.24

0.86–1.78

1.28

0.89–1.84

1.26

0.88–1.80

3

1.21

0.93–1.58

1.28

0.99–1.65

1.37

1.03–1.81

1.41

1.04–1.89

1.33

0.94–1.87

1.40

0.97–2.01

1.54

1.05–2.26

1.51

1.04–2.19

Least healthy Wald p value (N = 2,338)

\0.001

\0.001

\0.001

** Family factors: parental conflict and family social support

positive relationship to SDQ scores (ß = 0.78, 95 % CI 0.03–1.52), which was attenuated by adjustments for other health behaviours. There were no relationships observed between scores on either dietary measure and SMFQ scores at follow-up.

Discussion Key results In a cross-sectional study, we found evidence for an association between an Unhealthy diet and mental health problems in this sample of ethnically diverse and socially deprived adolescents; those in the highest quintile of Unhealthy diet score were more than twice as likely to be symptomatic on the SDQ, and nearly 50 % more likely to be symptomatic on the SMFQ after taking all identified confounders into account. There was also weaker evidence for an inverse association between a measure of Healthy diet and mental health problems. There was less evidence for a prospective association between measures of diet quality and case level mental health problems, although the patterns of associations between high scorers on both Healthy (inverse) and Unhealthy (positive) diet scales and caseness on the SDQ were the in the same direction as cross-sectional analyses. When continuous scores on the mental health questionnaires were examined as outcome measures, weak relationships between high scores on the Healthy diet scale and lower SDQ scores at follow-up were evident.

Strengths and limitations To our knowledge, this study is the first to report on the prospective relationship between diet quality and mental health in adolescents from highly disadvantaged and ethnically diverse backgrounds. An obvious strength of the study is the inclusion of a wide range of potentially confounding variables such as socio-economic status, family conflict and support, dieting behaviours, as well as drug use and other lifestyle behaviours. However, as in all such studies, we may not have adequately accounted for these factors, and residual confounding may be an explanation for these findings. Socio-economic status is a potentially important confounder in the relationship between diet quality and mental health, but we did not find that measures of socio-economic status confounded the relationships under investigation. This is likely to be a result of the relatively homogeneous nature of the study cohort with regards to social class, which is another strength of this study. A further strength of the study was the utilisation of two well-validated mental health scales for adolescents, capturing constructs of both depression and behaviour, and affording a replication of our previous Australian study using the SMFQ [7]. The main limitation relates to the Healthy diet scale in this study. Only three items were collected for this scale— fruit and vegetable intake and breakfast consumption. There may have been a particular problem for adolescents to adequately identify how many vegetables they consumed per day given the composition of meals such as

123

1304

curries and stir-fries, which are more likely to be consumed by adolescents from Asian and African backgrounds. Neither did the questions used to assess Healthy diet quality include several other accepted components of a healthy diet (such as whole grains, low fat dairy and fish). Moreover, we had no information on the quality of the breakfasts consumed in this cohort, where regular unhealthy food consumption was common and likely to be also reflected in breakfast choices. On the other hand, junk and processed foods, such as crisps, savoury snacks, sweets, biscuits, fried food, chips and soft drinks are ubiquitous in the community and their consumption less culturally determined. As such, the Unhealthy scale is likely to have adequately captured the consumption of these foods. Although the survey items used would have been less accurate than more in-depth measures of dietary intake, we have previously used simple dietary questionnaires to adequately rank students in this age group on their dietary quality [5, 7]. It is quite likely that statistical power was an issue in this study. In this cohort of socially disadvantaged adolescents, smoking, alcohol and drug use and family factors were more strongly related to the mental health outcomes than diet, and it is likely that any effect of diet on mental health would have been ‘washed out’ by the larger contribution of these other factors. The contention that statistical power was an issue in this study is supported by the stronger relationships observed in analyses utilising continuous outcomes rather than case level illness. The recognised limitations of the dietary scales may have also reduced statistical power and limited the ability to detect true associations. The fact that weak prospective relationships between Healthy diet scores and mental health were observed suggests that utilising larger samples may yield a clearer picture of the prospective relationship between diet and mental health in future studies. A response rate of 84 % is reasonable, and a number of strategies were used to encourage reliable response for an adolescent sample. The overall follow-up was 75 % and there was no difference in follow-up by Unhealthy diet scores. On the Healthy diet measures, however, those that scored very high or very low were less likely to be followed up (P = 0.04). This would have the effect of reducing variance, and thus statistical power. Other groups were also less likely to be followed up. These were: adolescents who were depressed at baseline, girls, those eligible for free school meals and white pupils. These patterns of follow-up may have had an impact on the associations found, and have contributed to an underestimation of the strength of association between diet quality and mental health at follow-up. There may have been some degree of social desirability bias in the answers to questions regarding the frequency of fruit and vegetable consumption, as

123

Soc Psychiatry Psychiatr Epidemiol (2013) 48:1297–1306

educational programs directed at this age group encourage the consumption of these foods. Moreover, given that both diet and mental health measures were reliant on self-reports by the adolescents, bias may have also resulted from differential dietary reporting by those with depressive symptoms; depressed adolescents may be more likely to report negative eating habits. Additionally, reverse causality may explain our results; mental health problems may have resulted in impaired eating habits and reduced self-care, with a concomitant reduction in dietary quality. As we did not have data on dietary intakes at follow-up, we were unable to test this hypothesis. Finally, unrecognised confounding cannot be ruled out as an explanation for these findings; we lacked data on variables such as maternal depression and personality factors, which may explain these demonstrated relationships. Interpretation The finding of a cross-sectional relationship between a measure of Unhealthy food intake and adolescent mental health in this study is concordant with our previous findings in Australian adolescents [5]. In the Australian study, participants were deliberately sampled from a wide range of socio-economic, geographical and demographic backgrounds which is in contrast to this study, wherein participants were drawn from a small area of East London with a similar sociodemographic profile. The fact that similar patterns of association between Unhealthy diets and mental health were seen in both studies, over and above the contribution made by a wide range of possible confounding factors, supports the veracity of this relationship. Another Australian study also reported positive associations between the consumption of takeaway foods and sweets and internalising and externalising behaviours in nearly 2,000 adolescents [6]. In that study, the strength of the association was stronger for externalising than internalising behaviours. This is concordant with our study where we consistently found stronger associations between diet and SDQ scores than with SMFQ scores. This is surprising, given that there were far fewer cases on the SDQ than on the SMFQ. This finding may reflect an impact of diet on behaviour more than depression and also indicate that boys, who tend towards external manifestations of emotional distress more than females [20], may be particularly affected by excessive consumption of poor quality foods. The lack of a strong evidence for a longitudinal association between diet quality and mental health does not mean that such a relationship does not exist; results of the longitudinal analyses, whilst not statistically significant, demonstrated similar patterns of association to the crosssectional analyses for high scorer on both scales, and were largely in the directions predicted by the hypothesis,

Soc Psychiatry Psychiatr Epidemiol (2013) 48:1297–1306

particularly for the healthy diet scale. Moreover, recent research in a large study of Spanish adults has demonstrated an inverse association between the level of adherence to a Mediterranean dietary pattern and the risk for incident depression [3], while the Whitehall II cohort study of middle-aged adults reported an increased risk for selfreported depression after 5 years for those adhering more strongly to an Unhealthy dietary pattern, and a reduced risk for those following a healthy dietary pattern [2]. Similarly, a recent prospective study of Australian adolescents demonstrated an independent contribution of diet quality to adolescent depression over a 2-year period [7]. In each of these prospective studies, the associations did not appear to be explained by reverse causality i.e. depression prompting dietary changes, or by confounding by sociodemographic factors or health behaviours. These recent data, therefore, support a causal role for diet quality in depressive illness. As discussed, any influence of diet on mental health is likely to be difficult to detect in the presence of a large number of potentially more powerful detrimental social, behavioural and environmental factors, and further studies in cohorts with increased social and other environmental problems may require larger sample sizes. It is also the case that diet is a long-term exposure with the potential to influence the development of the brain [21, 22] and the immune system [23] over a long period. Even peri-natal and early post-natal nutritional exposures may be of critical importance to later physical and mental health [24, 25]. As such, detecting an impact of dietary quality on mental health in adolescence may require cohort studies that examine the relationship from a much earlier period of life.

Conclusion Mindful of the aforementioned caveats, our findings are broadly in line with the association of diet quality with depression previously reported in Australian adolescents [5, 7], and offer some support for the contention that poor diet is a risk factor for mental health problems in adolescents. However, this is a nascent field of investigation, with further research necessary to confirm diet as a causal factor in depressive illnesses. Given the substantial burden of illness of adolescent depression and the fact that the majority of mental illnesses in the population manifest before age 25 [26], establishing whether or not poor quality diets are a risk factor for mental health problems in adolescents is of primary public health importance. As additional prospective cohort studies are warranted, further research should also incorporate intervention studies to examine whether dietary improvement results in improvements in mental health in both adults and adolescents.

1305 Acknowledgments We are grateful for the support of the schools, parents and students involved in this study, as well as the Community Advisory Board. We also thank the field team, including Wendy Isenwater, Giash Ahmed, Sarah Brentnall, Sultana Choudry-Dormer, Franca Davenport, Davina Woodley-Jones, Amanda Lawrence, Rachel Cameron and Hannah Bennett. Phase 1 of the RELACHS study was commissioned by the East London and the City Health Authority and Phase 2 by the Teenage Pregnancy Unit at the Department of Health: we thank them for their support. We also thank Tower Hamlets, City and Hackney and Newham Primary Care Trusts for additional funding. Associate Professor Jacka was the recipient of post-graduate scholarship funding from Australian Rotary Health and is supported by a NH&MRC Post-Doctoral Fellowship (#628912). Dr Rothon is funded by a Medical Research Council Special Training Fellowship (G0601707).

References 1. Jacka FN, Pasco JA, Mykletun A, Williams LJ, Hodge AM, O’Reilly SL, Nicholson GC, Kotowicz MA, Berk M (2010) Association between western and traditional diets and depression and anxiety in women. Am J Psychiatry 167:305–311 2. Akbaraly TN, Brunner EJ, Ferrie JE, Marmot MG, Kivimaki M, Singh-Manoux A (2009) Dietary pattern and depressive symptoms in middle age. Br J Psychiatry 195:408–413 3. Sanchez-Villegas A, Delgado-Rodriguez M, Alonso A, Schlatter J, Lahortiga F, Majem LS, Martinez-Gonzalez MA (2009) Association of the Mediterranean dietary pattern with the incidence of depression: the Seguimiento Universidad de Navarra/ University of Navarra follow-up (SUN) cohort. Arch Gen Psychiatry 66:1090–1098 4. Sanchez-Villegas A, Toledo E, de Irala J, Ruiz-Canela M, Pla-Vidal J, Martinez-Gonzalez MA (2012) Fast-food and commercial baked goods consumption and the risk of depression. Public Health Nutr 15:424–432 5. Jacka FN, Kremer PJ, Leslie E, Berk M, Patton G, Toumbourou JW, Williams JW (2010) Associations between diet quality and depressed mood in adolescents: results from the Healthy Neighbourhoods study. Aust N Z J Psychiatry 44:435–442 6. Oddy WH, Robinson M, Ambrosini GL, O’Sullivan TA, de Klerk NH, Beilin LJ, Silburn SR, Zubrick SR, Stanley FJ (2009) The association between dietary patterns and mental health in early adolescence. Prev Med 49:39–44 7. Jacka F, Kremer PJ, Berk M, de Silva-Sanigorski A, Moodie M, Leslie E, Pasco J, Swinburn B (2011) A prospective study of diet quality and mental health in adolescents. PLoS ONE 6:e24805 8. Stansfeld S, Haines MM, Booy R, Taylor SJ, Viner R, Head J (2003) Health of Young People in East London: The RELACHS Study 2001. The Institute of Community Health Sciences. Barts and The London Queen Mary’s School of Medicine & Dentistry, London 9. Stansfeld SA, Haines MM, Head JA, Bhui K, Viner R, Taylor SJ, Hillier S, Klineberg E, Booy R (2004) Ethnicity, social deprivation and psychological distress in adolescents: school-based epidemiological study in east London. Br J Psychiatry 185:233– 238 10. Wardle J, Sutton S, Jarvis M (1998) HABITS—The Health and Behaviours of Teenagers Study. Department of Epidemiology and Public Health, Health Behaviour Unit, London 11. Angold A, Costello EJ, Messer SC, Pickles A (1995) Development of a short questionnaire for use in epidemiological studies of depression in children and adolescents: factor composition and structure across development. Int J Meth Psych Res 5:237–249

123

1306 12. Goodman R, Meltzer H, Bailey V (1998) The Strengths and Difficulties Questionnaire: a pilot study on the validity of the selfreport version. Eur Child Adolesc Psychiatry 7:125–130 13. Meltzer H, Gatward R, Goodman R, Ford T (2003) Mental health of children and adolescents in Great Britain. Int Rev Psychiatry 15:185–187 14. Leavey G, Hollins K, King M, Barnes J, Papadopoulos C, Grayson K (2004) Psychological disorder amongst refugee and migrant schoolchildren in London. Soc Psychiatry Psychiatr Epidemiol 39:191–195 15. Mullick MS, Goodman R (2001) Questionnaire screening for mental health problems in Bangladeshi children: a preliminary study. Soc Psychiatry Psychiatr Epidemiol 36:94–99 16. Goddard E, Higgins V (1999) Smoking, drinking and drug use among young teenagers in 1998. National Centre for Social Research, London 17. Health Education Authority (1997) Young people and health: health behaviour in school-aged children. Health Education Authority, London 18. Zimet G, Dahlem N, Zimet S, Farley G (1988) The Multidimentional Scale of Perceived Social Support. J Pers Assess 52:30–41 19. Rothon C, Head J, Clark C, Klineberg E, Cattell V, Stansfeld S (2009) The impact of psychological distress on the educational achievement of adolescents at the end of compulsory education. Soc Psychiatry Psychiatr Epidemiol 44:421–427

123

Soc Psychiatry Psychiatr Epidemiol (2013) 48:1297–1306 20. Rutter M, Caspi A, Moffitt TE (2003) Using sex differences in psychopathology to study causal mechanisms: unifying issues and research strategies. J Child Psychol Psychiatry 44:1092–1115 21. Bryan J, Osendarp S, Hughes D, Calvaresi E, Baghurst K, van Klinken JW (2004) Nutrients for cognitive development in school-aged children. Nutr Rev 62:295–306 22. Sharma S, Zhuang Y, Gomez-Pinilla F (2012) High-fat diet transition reduces brain DHA levels associated with altered brain plasticity and behaviour. Sci Rep 2:431 23. Lopez-Garcia E, Schulze MB, Fung TT, Meigs JB, Rifai N, Manson JE, Hu FB (2004) Major dietary patterns are related to plasma concentrations of markers of inflammation and endothelial dysfunction. Am J Clin Nutr 80:1029–1035 24. Lucas A (2005) Long-term programming effects of early nutrition—implications for the preterm infant. J Perinatol 25(Suppl 2):S2–S6 25. Galler JR, Bryce CP, Waber D, Hock RS, Exner N, Eaglesfield D, Fitzmaurice G, Harrison R (2010) Early childhood malnutrition predicts depressive symptoms at ages 11–17. J Child Psychol Psychiatry 51:789–798 26. Kessler RC, Berglund P, Demler O, Jin R, Merikangas KR, Walters EE (2005) Lifetime prevalence and age-of-onset distributions of DSM-IV disorders in the National Comorbidity Survey Replication. Arch Gen Psychiatry 62:593–602