1

Depression and Risk of Type 2 Diabetes: the Potential Role of Metabolic Factors

Norbert Schmitz1, 2,3,4, PhD; Sonya S Deschênes1, 2, PhD; Rachel J Burns1, 2, PhD; Kimberley J. Smith5, PhD; Alain Lesage6, MD; Irene Strychar4,7, Ed.D; Rémi Rabasa-Lhoret4,7,8, MD, PhD; Cassandra Freitas2,3, MSc; Eva Graham2,3, MSc; Philip Awadalla9, PhD; JianLi Wang10, PhD

1 Department of Psychiatry, McGill University, Montreal, Quebec, Canada 2 Douglas Mental Health University Institute, Montreal, Quebec, Canada 3 Department of Epidemiology and Biostatistics, McGill University, Montreal, Quebec, Canada4 Montreal Diabetes Research Centre, Montreal, Quebec, Canada5Department of Life Sciences, Brunel University London, Uxbridge, Middlesex, UK6Centre de Recherche Fernand Seguin, Hôpital Louis-H. Lafontaine, University of Montreal, Montreal, QC, Canada7Department of Nutrition, Faculty of Medicine, University of Montreal, and the Centre de Recherche du Centre Hospitalier de l’Université de Montréal (CRCHUM), Montreal, QC, Canada8Institut de Recherches Cliniques de Montréal, Montréal, Québec, Canada9 Department of Paediatrics, Faculty of Medicine, University of Montreal, Montreal, QC, Canada10Departments of Psychiatry and Community Health Sciences, Faculty of Medicine, University of Calgary, Calgary, Canada.

Running title: Depression, Metabolic Factors and Diabetes

Address for correspondence:Norbert Schmitz, PhD Douglas Mental Health University Institute McGill University6875 LaSalle BoulevardMontreal, Quebec, H4H 1R3CanadaTel.:1-514-761-6131, ext. 3379 Fax: 1-514-888-4064 E-mail: norbert.schmitz@mcgill.ca

Number of words: 3491

1

2

ABSTRACT

The aim of the present study was to evaluate the interaction between depressive symptoms and

metabolic dysregulations as risk factors for type 2 diabetes. The sample was comprised of 2525

adults who participated in a baseline and a follow-up assessment over a 4.5 year period in the

Emotional Health and Wellbeing Study (EMHS) in Quebec, Canada. A two-way stratified

sampling design was employed, based on the presence of depressive symptoms and metabolic

dysregulation (obesity, elevated blood sugar, high blood pressure, high levels of triglycerides,

and decreased high-density lipoprotein). A total of 87 (3.5%) individuals developed diabetes.

Participants with both depressive symptoms and metabolic dysregulation had the highest risk of

diabetes (adjusted odds ratio=6.61, 95% CI: 4.86-9.01), compared to those without depressive

symptoms and metabolic dysregulation (reference group). The risk of diabetes in individuals

with depressive symptoms and without metabolic dysregulation did not differ from the reference

group (adjusted odds ratio=1.28, 95% CI: 0.81-2.03), whereas the adjusted odds ratio for those

with metabolic dysregulation and without depressive symptoms was 4.40 (95% CI: 3.42-5.67).

The Synergy Index (SI=1.52; 95% CI: 1.07-2.17) suggested that the combined effect of

depressive symptoms and metabolic dysregulation was greater than the sum of individual effects.

An interaction between depression and metabolic dysregulation was also suggested by a

structural equation model. Our study highlights the interaction between depressive symptoms

and metabolic dysregulation as a risk factor for type 2 diabetes. Early identification, monitoring,

and a comprehensive management approach of both conditions might be an important diabetes

prevention strategy.

2

3

INTRODUCTION

Type 2 diabetes mellitus is a progressive chronic disease that is increasing at epidemic rates. A

recent study found that the prevalence of diabetes worldwide has increased by 45% between

1990 and 20131. Mental health problems often co-occur in people with type 2 diabetes2. Notably,

three meta-analyses suggest that the risk of incident diabetes is 37% to 60% greater for

individuals with depressive symptoms than individuals without depressive symptoms3-5. The

mechanisms linking depression and type 2 diabetes are not well understood though it is likely

that behavioral and biological factors contribute to this association. For example, various

lifestyle-related behaviors such as smoking, physical inactivity, and poor diet are associated with

depression and can contribute to the development of prediabetes and type 2 diabetes6. Indeed,

health behaviors are important components of various diabetes risk scores7-10.

Metabolic dysregulation, often described by the clustering of three or more metabolic traits such

as high triglycerides, low high-density lipoprotein cholesterol, high blood pressure, abdominal

obesity and impaired glucose regulation, has been used as a method of identifying individuals at

increased risk of type 2 diabetes11.

Depression accompanied by metabolic dysregulations has been identified as a subtype of

depression12, 13 and has been labeled “metabolic depression”14, 15. It is possible that individuals

with comorbid metabolic dysregulations and depression might be at particular risk of developing

type 2 diabetes. For example, these individuals have more problems with lifestyle

modifications16 (motivation and adherence) than people with metabolic dysregulations only, or

they may have more biological and metabolic risk factors than individuals with depression only.

3

4

Although depression and metabolic factors have each been demonstrated to increase the risk of

type 2 diabetes, the extent to which the combination of these factors (i.e., metabolic depression)

increases risk of type 2 diabetes is unknown. Using data from a community sample of adults

aged 40 to 69 years without diabetes at baseline, our aim was to identify the association between

depressive symptoms, metabolic dysregulations and type 2 diabetes over 4.5 years while

controlling for other non-metabolic diabetes risk factors (family history of diabetes,

sociodemographic characteristics and lifestyle related behaviors). We hypothesized that

individuals with both depressive symptoms and metabolic dysregulations at baseline would have

a higher risk of diabetes than those with depressive symptoms only and those with metabolic

dysregulations only.

METHODS AND MATERIAL

Design/setting and participants

The Emotional Health and Wellbeing Study (EMHS) is a follow-up of 2525 adults who

participated in the CARTaGENE (CaG) study in the province of Quebec, Canada. A detailed

description of the CaG study is provided elsewhere17. Briefly, CaG recruited 20,004 participants

aged 40–69 years from four metropolitan areas between July of 2009 and October of 2010 where

detailed health, lifestyle and socio-demographic information, physiological measures and

biological samples were collected.

A stratified subsample of CaG participants who: agreed to be contacted for a follow up study, did

not have diabetes at baseline and had information on depressive symptoms were contacted and

4

5

invited to participate in the EMHS four to five years after the baseline assessment. A two-way

stratification approach based on the presence of depressive symptoms and metabolic

dysregulation was used. Depressive symptoms were defined as having a PHQ-9 summary score

of 6 and higher18, which includes mild, moderate and severe depressive symptoms. Classification

for metabolic dysregulation was based on current definitions of the metabolic syndrome, and was

defined as three or more metabolic risk factors19, including: elevated blood pressure (BP

>130/85mmHg or use of anti-hypertensive medication), impaired glycaemic control (HbA1c

between 5.7% and 6.5%), low high-density lipoprotein cholesterol (< 1.03 mmol l-1 in men and

<1.30 mmol l-1 in women), elevated triglycerides (> 1.7mmol l-1) and central obesity ( waist

circumference ≥ 102 cm (men), ≥ 88 cm (women)). EMHS participants were recruited from the

four groups (a) neither no/mild depressive symptoms nor metabolic dysregulation (n=9491 in

CAG database) (reference group); b) metabolic dysregulation without depressive symptoms

(n=5193 in CAG database); c) depressive symptoms without metabolic dysregulation (n=1635 in

CAG database); and d) presence of both depressive symptoms and metabolic dysregulation

(n=1179 in CAG database). All participants in groups c) and d) and random samples of groups a)

and b) were invited to participate. A detailed description of the sampling is provided in Figure 1.

Potential participants were initially contacted by email and followed up by mail and phone by the

CaG group. They were informed about the aims of the EMHS study and provided informed

consent (written or electronic consent on a secure website) if they were interested in being

contacted by EMHS investigators for a follow-up interview. Participants who agreed to be

contacted by the EMHS team were contacted via telephone between October 2014 and March

2015 by trained interviewers. All participants received a detailed explanation of the study and

were asked for verbal informed consent. The Douglas Mental Health University Institute Ethics

5

6

Board (EMHS investigators) and the St-Justine Hospital Research Ethics Board (CaG

investigators) approved the consent procedures and the study protocol.

Figure 1 here

Measures

Self-reported diabetes was the central outcome variable. It was measured with the self-report

question, “In the past 5 years, has a doctor told you that you have diabetes?”. Additional

information on diabetes was collected, including date of onset, current diabetes treatment (oral

medication, insulin treatment and diet) and tests for HbA1c levels. This information was used to

validate self-reported diabetes onset.

Depressive symptoms were measured with the Patient Health Questionnaire (PHQ-9) at baseline.

Individuals were asked to what extent they had experienced each of nine depressive symptoms

over the past two weeks. Possible responses range from “not at all (0)” to “on nearly every day

(3)”. A summary score was computed from the nine individual items, with higher scores

indicating greater depression symptom severity. Reliability and validity of the PHQ-9 have

indicated it has sound psychometric properties18.

Diagnostic assessments of 12-month and lifetime major depression disorder (MDD) were

conducted at follow-up using the computerized World Mental Health—Composite International

Diagnostic Interview (CIDI) 2.120, a standardized instrument for the assessment of mental

disorders according to the Diagnostic and Statistical Manual of Mental Disorders, Fourth

Edition, Text Revision21 criteria. The lifetime version of the CIDI collects information on age at

6

7

the most recent episode and age at first onset. Age at first onset was used to determine whether

an episode of depression had occurred before or after baseline assessment.

Physical activity at baseline was measured using the International Physical Activity

Questionnaire, an instrument which has shown to have good test-retest reliability and moderate

convergent validity with accelerometers22. Energy expenditure, measured as Metabolic

Equivalent (MET), was estimated23. Smoking status at baseline was assessed by two questions:

“Do you currently smoke cigarettes?” and “In your lifetime have you smoked a total of 100

cigarettes or more?” Smoking was categorized into three groups (current smoker, former smoker,

and never smoker). Participants were asked how many servings of vegetables (fresh, frozen,

canned or cooked) and fruit they eat on a typical day (serving was defined as ½ cup or 125mL).

A binary diet variable was defined as consuming five or more servings of fruits and vegetables

daily24.

Medication intake was assessed during the baseline interview. Antidepressant medications were

identified from the list of medications.

The CARTaGENE baseline survey collected data on socio-demographic characteristics.

Education was categorized into three categories (less than high school, high school, and

college/graduate studies/university). Ethnicity was classified into white vs non-white. Family

history of diabetes was assessed by asking questions about parents' and siblings' histories of

diabetes.

Analyses

In a first step we used the PHQ-9 for the assessment of depressive symptoms and compared

diabetes incidence for the four groups described above. Logistic regression models were used to

7

8

estimate the adjusted odds ratios and 95% confidence intervals of diabetes incidence for those

with a) neither no/mild depressive symptoms nor metabolic dysregulation (reference group); b)

metabolic dysregulation without depressive symptoms; c) depressive symptoms without

metabolic dysregulation; and d) presence of both depressive symptoms and metabolic

dysregulation.

Adjustment was made for age, sex, education, ethnicity, family history of diabetes, smoking, diet

and physical activity. All covariates were measured at baseline. We generated multiple imputed

values for the missing data from the variables used in the analysis (PROC MI and PROC

MIANALYSE in SAS 9.3). To account for the stratified sampling, we determined sampling

weights, defined as 1/sampling fraction, for each depression/metabolic dysregulation group and

incorporated them into the analyses.

Interaction between depressive symptoms and metabolic dysregulation was assessed as departure

from additive risks and quantified using Rothman’s synergy index (SI)25. SI is calculated as the

ratio between combined effect and individual effects: ORAB−1/[ORAb−1]+[ORaB−1] where AB is

the presence of both depressive symptoms and metabolic dysregulation, Ab is the presence of

depressive symptoms only and aB is the presence of metabolic dysregulation. A value greater

than 1 implies synergism (departure from additive risks)25.

In a second step we used structural equation modelling26 to determine the joint associations of

depression and metabolic risk factors on diabetes incidence. Structural equation models are an

extension of regression analysis that can handle measurement error and a wider range of

relationships among variables. Both depression and metabolic dysregulation were modeled as

latent variables. Those latent variables summarize the information measured with the observed

8

9

variables and account for measurement error and the individual contribution of each measure26.

Antidepressant medication, PHQ-9 summary score and depression history (based on CIDI

interview) at baseline were used as indicators for depression. Elevated blood pressure, impaired

glycaemic control, low high-density lipoprotein cholesterol, elevated triglycerides and central

obesity were used as indicators for the latent variable metabolic dysregulation. Diabetes

incidence was entered as a manifest variable. We controlled for the same baseline variables as in

our logistic regression analyses. We first specified a model without interaction between

depression and metabolic dysregulation, followed by a model that included an interaction term

between the two latent variables (multiplicative interaction). Models were compared for fit using

the Akaike Information Criterion (AIC) and the Sample-Size Adjusted Bayesian Information

Criterion (BIC) and a Wald test was used for model comparison27. Parameters were estimated

using a maximum likelihood estimation, including estimating a random effect and maximum

likelihood estimator with robust standard errors. The estimated path coefficients from the latent

variables to the diabetes incidence variable can be interpreted as odds ratios. Analyses were

conducted with MPlus (Version 7.3)28.

Code availability

All relevant computer code, necessary to reproduce our results, can be obtained by the authors

upon request.

RESULTS

A total of 2,525 individuals participated in the follow-up assessment. Sociodemographic and

clinical characteristics of participants in our cohort and the original CARTaGENE cohort were

9

10

similar, although EMHS participants had a somewhat higher educational level. Result are

presented in Table 1. Mean follow-up time was 4.6 years (SD=0.30). The mean age at baseline

was 53.9 (SD=7.5) years and the proportion of women was 57.4%. Baseline characteristics are

presented in Table 2.

Table 1 here

Table 2 here

There were 760 (30.1%) participants in group a) (no/mild depressive symptoms and no metabolic

dysregulation) (reference group); 734 (29.1%) participants in group b) (metabolic dysregulation

without depression); 595 (23.6%) participants in group c) (depressive symptoms without

metabolic dysregulation); and 436 (17.3%) participants in group d) (presence of both depressive

symptoms and metabolic dysregulation). Participants with metabolic dysregulation and

depressive symptoms (group d) had a higher waist circumference (p<0.001) but did not have a

worse metabolic dysregulation (diastolic blood pressure, HbA1c levels, HDL levels and

triglyceride levels) than participants with metabolic dysregulation without depressive symptoms

(group b).

A total of 87 (3.5%) individuals developed diabetes between the baseline assessment and the

follow-up assessment. Diabetes incidence in the four groups was 0.9%, 4,6%, 1.5% and 8.5%,

respectively.

10

11

Table 2 shows results that were obtained from unadjusted and adjusted logistic regression

analyses. Those without depression and without metabolic dysregulation at baseline were defined

as the reference category. Metabolic dysregulation without depressive symptoms was associated

with an increased risk for diabetes incidence in both unadjusted (OR=5.23) and adjusted

(OR=4.40) models. Depressive symptoms without metabolic dysregulation were no longer

significantly associated with diabetes incidence when controlling for other diabetes risk factors.

Participants with both depressive symptoms and metabolic dysregulation had the highest risk for

diabetes incidence. The Synergy Index for the adjusted model was 1.52 (95% CI: 1.07 to 2.17),

suggesting that the combined effect of depressive symptoms and metabolic dysregulation is

greater than the sum of the individual effects.

In sensitivity analyses, we excluded participants with high waist circumference (top percentile of

the empirical distribution: ≥ 120 cm (men) and ≥ 108 cm (women)) to ensure comparable waist

circumference for those with metabolic dysregulation and depressive symptoms (i.e., group d)

and those with metabolic dysregulation without depressive symptoms (i.e., group b). This did not

provide substantially different results from our main analyses (data not shown).

Table 3 here

The structural equation model containing the interaction between the latent variables

representing depression and metabolic dysregulation had better fit than the model without the

interaction. The Wald tests also indicated better fit for the model with interaction (p<0.05). The

11

12

structural equation model with the latent variables interaction and corresponding path

coefficients and odds ratios is presented in Figure 2.

For the measurement portion of the model, when setting impaired glycaemic control as the

reference, all metabolic indicators were significant for the latent variable metabolic

dysregulation. When setting antidepressant medication as the reference, both depressive

symptoms (PHQ-9) and major depression history (CIDI interview) were significant for the latent

variable depression. Path coefficients between the latent variable and the indicators can be

interpreted as regression coefficients. For the structural equation model there was a significant

depression-metabolic dysregulation association (OR=1.59, 95% CI: 1.05-2.41) and a significant

main effect for metabolic dysregulation but not for depression.

Figure 2 here

DISCUSSION

In this prospective community study of 2525 individuals aged 40 to 69 years without type 2

diabetes at baseline, we evaluated the impact of depression and metabolic dysregulations on type

2 diabetes incidence over approximately 4.5 years. The results suggest an interaction between

these two factors in relation to an increased risk of type 2 diabetes. Logistic regression analyses

suggested that individuals with both depressive symptoms and metabolic dysregulations at

baseline were at higher risk for developing diabetes than participants with either one of the

conditions only. The combined effect of the two conditions appeared to be more than the sum of

12

13

the individual effects, as suggested by the synergy index (additive interaction). It has been

emphasized that measuring effects on the additive scale is most appropriate for prevention and

for assessing the public health relevance of an exposure, since the additive scale indicates

whether the effect of a risk factor would be greater in one subpopulation than in another29.

Departure form additivity is also used to study biological interaction, which is defined as two

causes acting in the same sufficient-component model to cause disease30.

A multiplicative interaction between depression and metabolic dysregulation was also suggested

by a structural equation model where depression was defined based on a clinical interview, a

symptom scale and antidepressant medication. The structural equation model takes measurement

error into account and allows a more flexible modelling.

To our knowledge, this is the first prospective community study designed to estimate the joint

effect of depression and metabolic dysregulations on incident type 2 diabetes in individuals aged

40 to 69 years.

The main strengths of the study are the prospective design, the availability of metabolic factors at

baseline, a clinical interview for the assessment of depression history and the stratified sampling.

Strengths of the analysis include a structural equation modeling approach. This method has

several advantages: a) depression and metabolic dysregulation were modeled as continuous

latent constructs rather than categorized summary scores, which reduces potential

misclassification associated with cutoff point for binary or categorical variables; b) we assumed

that latent constructs were measured with measurement error; that is, depression and metabolic

dysregulation were not assumed to be perfectly measured by the indicators; c) different

indicators for the two latent constructs had different weights (estimated coefficients in Figure 2);

13

14

and d) all parameters were estimated simultaneously in one model. The structural equation

approach permitted consideration of a depression construct that is more general than depressive

symptoms in the past two weeks, as measured by the PHQ-9; we modeled depression as a latent

variable where current depressive symptoms, antidepressant medication and major depression

history based on a structured diagnostic interview were included as indicators for the depression

construct.

We acknowledge several study limitations. Diabetes incidence was based on self-reports and not

on clinical measures. Individuals with undiagnosed diabetes were not identified. There is the

possibility that some individuals with undiagnosed diabetes might have higher levels of

depressive symptoms. Nonetheless, evidence from clinical and population based studies suggests

that self-reported diabetes is a robust exposure measure31, 32 and in general a sufficiently accurate

measure of diabetes in epidemiologic studies33. Depressive symptoms at baseline were assessed

using a brief self-report scale that measures depressive symptoms experienced in the past two

week and does not account for history and treatment of depression. A clinical interview for

depression (history) was conducted at follow-up assessment only and might be subject to recall

bias. In addition, lifestyle-related behaviours such as smoking, diet and physical activity were

assessed by self-report, which may be subjected to reporting bias. We cannot exclude the

possibility that selection bias might have affected the strength of the association. Awadalla et al17

found an overall concordance in the distribution of socio-demographic characteristics between

the CaG cohort and the general population, although CaG participants were generally more

educated. Furthermore, those people who took part in the CaG study were likely healthier than

the overall cohort due to restrictions placed upon eligibility for blood assessments. Finally, as in

14

15

all observational studies, there could be unmeasured confounding by unknown or unmeasured

predictors.

There are several ways in which depression or depressive symptoms may be associated with an

increased risk of developing type 2 diabetes. Metabolic factors might play a key role in this

association. Metabolic dysregulations are important and well established risk factors for diabetes

and require a variety of self-management behaviors to optimize treatment. The management of

metabolic dysregulations often includes increasing physical activity, changing one’s diet, and

smoking cessation. Depression has been shown to adversely impact these self-management

behaviors16, which might worsen the management of metabolic factors. Additionally, depression

has been shown to be associated with poor adherence to medication34. It is likely that depressive

symptoms and metabolic dysregulations interact with each other in a dynamic way: depressive

symptoms can impede people from managing their metabolic conditions as effectively as they

need to, which can lead to more metabolic problems, which, in turn, can result in more

depressive symptoms35.

Metabolic dysregulations and depression also share common pathophysiologic mechanisms36 and

the co-occurrence of both conditions might amplify the risk of developing diabetes. Depression

is associated with sympathetic nervous system activation and dysregulation of the hypothalamic-

pituitary-adrenocortical (HPA) axis37), which are associated with metabolic changes through

abdominal fat accumulation, glucose metabolism, and blood pressure regulation. Some

antidepressant medications can cause weight gain and obesity38 and are associated with an

increased risk of diabetes38. Increased levels of inflammatory cytokines (e.g., C-reactive protein

and interleukin 6)39 are also associated with both metabolic dysregulation and depression.

Depression and metabolic dysregulation might stimulate each other’s occurrence, which can in

15

16

turn result in several metabolic dysregulations. This can lead to a vicious cycle that further

aggravates depression and metabolic outcomes14. For example, Vogelzangs et al.40 have shown

that metabolic dysregulations predicted a more chronic course of depressive disorders. Recurrent

or chronic depressive symptoms are associated with prolonged exposure to psychosocial and

normal life stress, which can wear and tear different circulatory, inflammatory, immune and

psychological regulatory systems41, 42 and increase allostatic load (the price the body pays for

being forced to adapt to adverse psychosocial or physical situations)41.

Depression is a heterogeneous condition. Data-driven techniques have confirmed that depression

can generally be divided into several subtypes43. Lamers et al.44 identified three depressive

subtypes within a cohort of subjects with depression using latent class analysis (severe

melancholic class, 46%; severe atypical class, 25%; and moderate severity, 29%). The severe

atypical class was associated with more metabolic dysregulations, suggesting that that this

subtype involves a metabolic component. Depression subtypes with metabolic components have

also been reported by others12, 45.

Although many studies and meta-analyses have shown that depression is associated with

increased diabetes incidence, depression alone might not be the focal point of the problem, but

rather the combination of depression with metabolic dysregulations. It might be crucial for a

better understanding of depression as risk factor for diabetes and for future prevention

interventions to identify and focus on homogeneous depression phenotypes that take biological

correlates into account.

CONCLUSIONS

16

17

In conclusion, our study conducted in a community sample highlights the interaction between

depressive symptoms and metabolic dysregulation as a potentially important risk factor for type

2 diabetes. Early identification, monitoring, and a comprehensive management approach of both

conditions might be an important diabetes prevention strategy.

17

18

Acknowledgment:

Author Contributions: This work was supported by Operating Grant MOP-130552 from the

Canadian Institutes of Health Research (CIHR). The funding agencies had no role in the design

or conduct of the study, in the collection, management, analysis, or interpretation of the data, or

in the preparation, review, or approval of the manuscript. Sonya Deschênes is supported by a

fellowship from the Fonds de recherche du Québec – Santé, Canada and Rachel Burns is

supported by a fellowship from the Canadian Institutes of Health Research.

Conflict of Interest Disclosures: The authors have no conflicts of interest.

18

19

REFERENCES

1. Collaborators GBoDS. Global, regional, and national incidence, prevalence, and years lived with disability for 301 acute and chronic diseases and injuries in 188 countries, 1990-2013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet 2015.

2. Ducat L, Philipson LH, Anderson BJ. The Mental Health Comorbidities of Diabetes. Jama-Journal of the American Medical Association 2014; 312(7): 691-692.

3. Knol MJ, Twisk JWR, Beekman ATF, Heine RJ, Snoek FJ, Pouwer F. Depression as a risk factor for the onset of type 2 diabetes mellitus. A meta-analysis. Diabetologia 2006; 49(5): 837-845.

4. Mezuk B, Eaton WW, Albrecht S, Golden SH. Depression and Type 2 Diabetes Over the Lifespan A meta-analysis. Diabetes Care 2008; 31(12): 2383-2390.

5. Rotella F, Mannucci E. Depression as a Risk Factor for Diabetes: A Meta-Analysis of Longitudinal Studies. J Clin Psychiatry 2013; 74(1): 32-38.

6. Renn BN, Feliciano L, Segal DL. The bidirectional relationship of depression and diabetes: A systematic review. Clin Psychol Rev 2011; 31(8): 1239-1246.

7. Schulze MB, Hoffmann K, Boeing H, Linseisen J, Rohrmann S, Mohlig M et al. An accurate risk score based on anthropometric, dietary, and lifestyle factors to predict the development of type 2 diabetes. Diabetes Care 2007; 30(3): 510-515.

8. Lindstrom J, Tuomilehto J. The diabetes risk score - A practical tool to predict type 2 diabetes risk. Diabetes Care 2003; 26(3): 725-731.

9. Rahman M, Simmons RK, Harding AH, Wareham NJ, Griffin SJ. A simple risk score identifies individuals at high risk of developing Type 2 diabetes: a prospective cohort study. Fam Pr 2008; 25(3): 191-196.

10. Joseph J, Svartberg J, Njolstad I, Schirmer H. Incidence of and risk factors for type-2 diabetes in a general population: The Tromso Study. Scandinavian Journal of Public Health 2010; 38(7): 768-775.

11. Shin JA, Lee JH, Lim SY, Ha HS, Kwon HS, Park YM et al. Metabolic syndrome as a predictor of type2 diabetes, and its clinical interpretations and usefulness. J Diabetes Investig 2013; 4(4): 334-343.

19

20

12. McIntyre RS, Rasgon NL, Kemp DE, Nguyen HT, Law CWY, Taylor VH et al. Metabolic syndrome and major depressive disorder: Co-occurrence and pathophysiologic overlap. Curr Diabetes Rep 2009; 9(1): 51-59.

13. Vancampfort D, Correll CU, Wampers M, Sienaert P, Mitchell AJ, De Herdt A et al. Metabolic syndrome and metabolic abnormalities in patients with major depressive disorder: a meta-analysis of prevalences and moderating variables. Psychol Med 2014; 44(10): 2017-2028.

14. Vogelzangs N, Beekman ATF, Boelhouwer IG, Bandinelli S, Milaneschi Y, Ferrucci L et al. Metabolic Depression: A Chronic Depressive Subtype? Findings From the InCHIANTI Study of Older Persons. J Clin Psychiatry 2011; 72(5): 598-604.

15. Liu CS, Carvalho AF, McIntyre RS. Towards a "Metabolic" Subtype of Major Depressive Disorder: Shared Pathophysiological Mechanisms May Contribute to Cognitive Dysfunction. Cns & Neurological Disorders-Drug Targets 2014; 13(10): 1693-1707.

16. DiMatteo MR, Lepper HS, Croghan TW. Depression is a risk factor for noncompliance with medical treatment - Meta-analysis of the effects of anxiety and depression on patient adherence. Archives of Internal Medicine 2000; 160(14): 2101-2107.

17. Awadalla P, Boileau C, Payette Y, Idaghdour Y, Goulet JP, Knoppers B et al. Cohort profile of the CARTaGENE study: Quebec's population-based biobank for public health and personalized genomics. Int J Epidemiol 2013; 42(5): 1285-1299.

18. Lamers F, Jonkers CCM, Bosma H, Penninx B, Knottnerus JA, van Eijk JTM. Summed score of the Patient Health Questionnaire-9 was a reliable and valid method for depression screening in chronically ill elderly patients. Journal of Clinical Epidemiology 2008; 61(7): 679-687.

19. Alberti KGMM, Eckel RH, Grundy SM, Zimmet PZ, Cleeman JI, Donato KA et al. Harmonizing the Metabolic Syndrome A Joint Interim Statement of the International Diabetes Federation Task Force on Epidemiology and Prevention; National Heart, Lung, and Blood Institute; American Heart Association; World Heart Federation; International Atherosclerosis Society; and International Association for the Study of Obesity. Circulation 2009; 120(16): 1640-1645.

20. Kessler RC, Ustun TB. The World Mental Health (WMH) Survey Initiative Version of the World Health Organization (WHO) Composite International Diagnostic Interview (CIDI). International journal of methods in psychiatric research 2004; 13(2): 93-121.

21. American Psychiatric Association APATFoDSMIV. Diagnostic and statistical manual of mental disorders : DSM-IV-TR. American Psychiatric Association: Washington, DC, 2000.

20

21

22. Craig CL, Marshall AL, Sjostrom M, Bauman AE, Booth ML, Ainsworth BE et al. International physical activity questionnaire: 12-country reliability and validity. Med Sci Sports Exerc 2003; 35(8): 1381-1395.

23. Guidelines for data processing and analysis of the International Physical Activity Questionnaire (IPAQ)–short and long forms. https://sites.google.com/site/theipaq/scoring-protocol, 2005, Accessed Date Accessed 2005 Accessed.

24. Eyre H, Kahn R, Robertson RM, Writing AAAC. Preventing cancer, cardiovascular disease, and diabetes - A common agenda for the American Cancer Society, the American Diabetes Association, and the American Heart Association. Diabetes Care 2004; 27(7): 1812-1824.

25. Andersson T, Alfredsson L, Kallberg H, Zdravkovic S, Ahlbom A. Calculating measures of biological interaction. Eur J Epidemiol 2005; 20(7): 575-579.

26. Bollen KA. Structural equations with latent variables. Wiley: New York, NY, 1989.

27. B B. Structural Equation Modeling with Mplus: Basic Concepts, Applications, and Programming. Routledge Academic Publishers: New York, 2012.

28. Muthén LKaM, B.O. Mplus User’s Guide. Seventh Edition. Muthén & Muthén: Los Angeles, CA:, 2012.

29. Knol MJ, VanderWeele TJ. Recommendations for presenting analyses of effect modification and interaction. International Journal of Epidemiology 2012; 41(2): 514-520.

30. Qin L, Knol MJ, Corpeleijn E, Stolk RP. Does physical activity modify the risk of obesity for type 2 diabetes: a review of epidemiological data. European Journal of Epidemiology 2010; 25(1): 5-12.

31. Tisnado DM, Adams JL, Liu HH, Damberg CL, Chen WP, Hu FA et al. What is the concordance between the medical record and patient self-report as data sources for ambulatory care? Medical Care 2006; 44(2): 132-140.

32. Skinner KM, Miller DR, Lincoln E, Lee A, Kazis LE. Concordance between respondent self-reports and medical records for chronic conditions: experience from the Veterans Health Study. The Journal of ambulatory care management 2005; 28(2): 102-110.

21

22

33. Margolis KL, Qi LH, Brzyski R, Bonds DE, Howard BV, Kempoinen S et al. Validity of diabetes self-reports in the Women's Health Initiative: comparison with medication inventories and fasting glucose measurements. Clinical Trials 2008; 5(3): 240-247.

34. Grenard JL, Munjas BA, Adams JL, Suttorp M, Maglione M, McGlynn EA et al. Depression and Medication Adherence in the Treatment of Chronic Diseases in the United States: A Meta-Analysis. J Gen Intern Med 2011; 26(10): 1175-1182.

35. Schmitz N, Gariepy G, Smith KJ, Clyde M, Malla A, Boyer R et al. Recurrent Subthreshold Depression in Type 2 Diabetes: An Important Risk Factor for Poor Health Outcomes. Diabetes Care 2014; 37(4): 970-978.

36. Marazziti D, Rutigliano G, Baroni S, Landi P, Dell'Osso L. Metabolic syndrome and major depression. CNS Spectr 2014; 19(4): 293-304.

37. Musselman DL, Evans DL, Nemeroff CB. The relationship of depression to cardiovascular disease - Epidemiology, biology, and treatment. Archives of General Psychiatry 1998; 55(7): 580-592.

38. Pan A, Sun Q, Okereke OI, Rexrode KM, Rubin RR, Lucas M et al. Use of antidepressant medication and risk of type 2 diabetes: results from three cohorts of US adults. Diabetologia 2012; 55(1): 63-72.

39. Eckel RH, Alberti K, Grundy SM, Zimmet PZ. The metabolic syndrome. Lancet 2010; 375(9710): 181-183.

40. Vogelzangs N, Beekman ATF, Dortland A, Schoevers RA, Giltay EJ, de Jonge P et al. Inflammatory and Metabolic Dysregulation and the 2-Year Course of Depressive Disorders in Antidepressant Users. Neuropsychopharmacology 2014; 39(7): 1624-1634.

41. McEwen BS. Mood disorders and allostatic load. Biol Psychiatry 2003; 54(3): 200-207.

42. Lesage A. Heuristic model of depressive disorders as systemic chronic disease. Epidemiology and psychiatric sciences 2015; 24(4): 309-311.

43. Penninx BWJH, Milaneschi Y, Lamers F, Vogelzangs N. Understanding the somatic consequences of depression: biological mechanisms and the role of depression symptom profile. Bmc Medicine 2013; 11.

22

23

44. Lamers F, de Jonge P, Nolen WA, Smit JH, Zitman FG, Beekman ATF et al. Identifying Depressive Subtypes in a Large Cohort Study: Results From the Netherlands Study of Depression and Anxiety (NESDA). J Clin Psychiatry 2010; 71(12): 1582-1589.

45. Cizza G, Ronsaville DS, Kleitz H, Eskandari F, Mistry S, Torvik S et al. Clinical Subtypes of Depression Are Associated with Specific Metabolic Parameters and Circadian Endocrine Profiles in Women: The Power Study. PLoS One 2012; 7(1): 9.

23

24

Table 1: Baseline characteristics of the original CARTaGENE cohort and the EMHS cohort, stratified by group status

Group a:no/mild depressive

symptoms and no metabolic dysregulation

Group bno/mild depressive

symptoms and metabolic dysregulation

Group cdepressive symptoms and no

metabolic dysregulation

Group ddepressive symptoms and metabolic dysregulation

BaselineN=9059

EMHS cohortN=760

BaselineN=5193

EMHS cohortN=734

BaselineN=1564

EMHS cohortN=595

BaselineN=1179

EMHS cohortN=436

Age, % 40 to 49 y 50 to 59 y 60 to 70 y

40.736.922.4

39.337.822.9

27.339.533.3

25.341.932.7

44.241.014.8

43.044.212.8

32.542.325.2

29.945.924.1

Sex, % Women 54.3 61.0 44.5 46.1 62.9 65.0 55.8 59.6Education, % Less than high school High school College/graduate studies/university

1.318.979.8

0.312.487.3

2.325.871.9

0.316.583.2

2.226.970.7

1.022.176.9

5.433.461.3

3.726.070.3

Metabolic risk factors Hypertension, % Impaired glycaemic control, % Low high-density lipoprotein cholesterol, % Elevated triglycerides, % Central obesity, %

Systolic blood pressure, M (SD) Diastolic blood pressure, M (SD) HbA1c (%), M (SD) HDL (mmol l-1), men, M (SD) HDL (mmol l-1), women, M (SD) Triglyceride (mmol l-1), M (SD) Waist circumf. (cm), men, M (SD) Waist circumf. (cm), women M (SD)

26.624.623.6

23.315.7

25.626.421.6

20.515.9

119.8 (14.4)71.2 (9.4)5.5 (0.3)

1.24 (0.35)1.62 (0.40)1.33 (0.78)92.9 (8.9)81.3 (9.7)

71.060.178.2

78.472.1

72.261.677.5

79.270.0

129.6 (14.4)77.0 (9.6)5.7 (0.3)

0.92 (0.24)1.15 (0.28)2.47 (1.23)

104. 8 (11.2)96.2 (11.1)

23.724.227.5

23.621.2

20.825.128.5

23.522.9

116.6 (13.8)70.2 (9.6)5.5 (0.3)

1.20 (0.33)1.50 (0.41)1.41 (0.86)93.4 (10.6)83.4 (11.5)

67.061.980.0

79.480.8

67.761.677.5

82.183.0

126.5 (14.4)77.0 (10.0)5.7 (0.3)

0.91 (0.24)1.12 (0.30)2.59 (1.36)109.1 (12.6)100.4 (11.5)

Moderate to severe depressive symptoms (PHQ 9≥ 11), % 0 0 0 0 28.5 24.9 27.9 24.8

24

25

Table 2: Baseline characteristics of EMHS participants, stratified by diabetes status at follow-up assessment

Participants who did not develop diabetes

N=2438

Participants who developed diabetesN=87

Age, M (SD) 53.9 (7.5) 54.0 (7.0)

Sex Women, % 57.7 49.4

Education Less than high school, % High school, % College/graduate studies/university, %

1.017.881.2

2.328.769.0

Ethnicity White, % 94.9 88.5

Diabetes Family history, % 38.5 60.5

Smoking Current smoker, % Former smoker, % Never smoker, %

16.740.043.3

27.643.728.7

Diet Five or more servings of fruits and vegetables daily, %

45.6 58.6

Physical activity Low, % Moderate, % High, %

17.440.342.3

18.545.735.8

Depression PHQ-9 Score, M (SD) 4.6 (4.6) 6.6 (5.7)

Depression Mild to severe symptoms (PHQ-9 ≥ 6), %

40.4 52.9

Major Depression history (CIDI interview), % 26.8 31.0

Antidepressant medication, % 15.3 20.7

Metabolic risk factors Hypertension, % Impaired glycaemic control, % Low high-density lipoprotein cholesterol, % Adverse triglycerides, % Central obesity, %

44.741.348.548.743.8

63.277.972.472.474.1

25

26

Table 3: Unadjusted and adjusted odds ratios of self-reported diabetes-onset by depressive symptoms/metabolic dysregulation categories in EMHS participants

Odds ratio for diabetes onset

(95% CI)

Adjusted odds ratiofor diabetes onset

(95% CI)

Baseline

Group a: no/mild depressive symptoms and no metabolic dysregulation

1 1

Group b: no/mild depressive symptoms and metabolic dysregulation

5.23(4.08 to 6.69)

4.40(3.42 to 5.67)

Group c: depressive symptoms and no metabolic dysregulation

1.65(1.05 to 2.59)

1.28(0.81 to 2.03)

Group d: depressive symptoms and metabolic dysregulation

9.98(7.44 to 13.38)

6.61(4.86 to 9.01)

26

27

Figure 1: Recruitment flow chart, EMHS Study, 209-2015

CARTaGENE Database (2009/2010) Group a Group b Group c Group dn= 9059 n= 5193 n=1564 n= 1179

Invitation to be contacted n= 2954 n= 2301 n= 1564 n= 1179for EMHS(random selectionfor groups a and b)

Response n= 871 n= 898 n= 908 n= 707

Consent tobe contacted n= 828 n= 794 n= 659 n= 485for EMHS2014

EMHScontacted and n= 760 n= 734 n= 595 n= 436interviewed2014-2015

Notes. EMHS participant flow chart. Group a = no depressive symptoms/no metabolic dysregulations; Group b = no depressive symptoms/metabolic dysregulations; Group c = depressive symptoms/no metabolic dysregulations; Group d4 = depressive symptoms/ metabolic dysregulations. All participants were contacted by email or mail and non-responder in groups c and d were followed up by email, mail and phone.

27

28

Figure 2: Structural equation model of relations between depression, metabolic dysregulation and diabetes incidence.

28

0.13 (0.02-0.24)

2.93 (1.99-3.88)

3.81 (2.50-5.17)

1.68 (1.12-2.25)

0.39 (0.22-0.55)

3.71 (2.37-5.06)

0.83 (0.48-1.17)

CovariatesSexAge Fam History DiabetSmokingDietPhysical ActivityEducationEthnicity

1.59 (1.05-2.41)

13.38 (4.51-39.70)

0.91 (0.74-1.11)

1 (Reference)

1 (Reference)

DiabetesIncidence

Metabolic Dysregulation

Depression

Central obesity

Adverse triglycerides

Adverse high-densitylipoprotein cholesterol

Impaired glycaemiccontrol

Elevated blood pressure

Current Antidepressant Medication

Major Depression History (CIDI interview)

Depressive Symptoms(PHQ-9 score)

29

Note: Rectangles represent measured variables, and circles represent latent constructs. AIC, BIC, and sample-size adjusted BIC for the models with and without interaction were 36970, 37158, 37053, 37260, 37244 and 37233, respectively)

29