This thesis is presented for the degree of. Doctor of Philosophy. of The University of Western Australia. Sunil K Bhat. MBBS MD MPH (research)

Transcription

1 Family functioning and related psychosocial factors in pregnancy and early childhood as determinants of cardiovascular risk factors in a longitudinal Australian pregnancy cohort: The Raine Study This thesis is presented for the degree of Doctor of Philosophy of The University of Western Australia by Sunil K Bhat MBBS MD MPH (research) School of Medicine and Pharmacology Royal Perth Hospital Unit 2017

2 Thesis Declaration I, Sunil Bhat, certify that: This thesis has been substantially accomplished during enrolment in the degree. This thesis does not contain material which has been accepted for the award of any other degree or diploma in my name, in any university or other tertiary institution. No part of this work will, in the future, be used in a submission in my name, for any other degree or diploma in any university or other tertiary institution without the prior approval of The University of Western Australia and where applicable, any partner institution responsible for the joint-award of this degree. This thesis does not contain any material previously published or written by another person, except where due reference has been made in the text. The work(s) are not in any way a violation or infringement of any copyright, trademark, patent, or other rights whatsoever of any person. This thesis contains published work and/or work prepared for publication, some of which has been co-authored. Sunil K Bhat Signature: Date: 19/06/2017 ii

3 Authorship Declaration iii

4 Abstract The overall aim of the thesis was to assess the influence of antenatal and postnatal psychosocial and familial determinants that may partly explain the relationships between young adulthood blood pressure (BP) and body mass index (BMI), prenatal stress and negative emotional states of depression and anxiety. The three studies undertaken in this thesis examined data from approximately 1000 young adults participating in the Western Australian Pregnancy Cohort (Raine) Study, a population based pregnancy cohort. The participants were the 20-year old offspring of pregnant women recruited between May 1989 and November Various environmental stressors in pregnancy have been reported to affect high BP and obesity in adults. Three contemporary pregnancy-cohort studies have examined offspring BP during childhood or adolescence, but to my knowledge, no study has examined the effect of prenatal maternal stress exposure on offspring BP and BMI in early adulthood. The first study aimed to assess the role of prenatal life stress exposure on offspring BP and BMI at age 20. The results showed that each additional prenatal life stress score predicted an increase in BMI by 0.37 kg/m 2 at age 20. In contrast, each additional prenatal life stress score reduced systolic BP by 0.66 mmhg in those with an average BMI of 24.5 kg/m 2. The relationship between prenatal life stress and BP was influenced by an interaction between prenatal life stress and offspring adiposity in such a way that young adults with a higher BMI showed an accentuation of this inverse association. Prenatal life stress exposure alone does not explain the rising trend of obesity among young adults. An increase in mental health disorders in young populations also contributes, to some extent, to the rise in obesity among young adults. While the literature suggests the theoretical possibility of a disadvantaged family environment influencing the association between depression and adiposity, no study has disentangled the underlying influences on this relationship. The second study therefore aimed to examine the complex relationship between offspring depression symptoms and BMI in the context of psychosocial and familial determinants. The results showed a positive relationship between depression symptoms and BMI. This association was significantly stronger among offspring of mothers that smoked during pregnancy or had a low family income status at pregnancy, the two being proxy indicators for an adverse family iv

5 environment. For offspring of maternal prenatal smokers, the positive association equated to a 1.1 kg/m 2 increase in BMI for every standard deviation (SD=8 units) increase in depression score. Depression and anxiety disorders are both psychosocial risk factors for cardiovascular disease (CVD). Studies that have examined the associations between BP and depression symptoms or clinical depression, have reported varied results (positive, inverse, or equivocal). The most common reasons for differing results are age-related comorbidities, bio-behavioural and lifestyle confounders; varying means of assessing depression or anxiety and heterogeneity of different BP outcome measures. The third study aimed to assess the relationships between BP and depression or anxiety symptoms, or self-reported history of depression at age 20, before the onset of confounding effects of major co-morbidity. The results showed an inverse relationship between depression symptoms and BP, which equated to a systolic BP (SBP) decrease of 0.79 mmhg for each additional SD increase in depression score. There was also a significant interaction between adiposity and the association between self-reported clinical depression and offspring blood pressure such that increasing BMI accentuated the inverse relationship between self-reported depression and SBP. The data presented in this thesis show that prenatal exposure to life stress or later depression or anxiety symptoms associate with lower resting BP and increased BMI in young adults. Possible mechanisms for these effects are presented. The relationship between depression symptoms at age 20 and young adult BMI is seen predominantly in those with a history of prenatal smoking and low family income, highlighting lower socio-economic groups most suitable for targeting preventative health measures. The results further show that the relationship between self-reported history of depression and lower BP at age 20 is accentuated by offspring adiposity. Explanations are presented to explain the apparent contrast between these latter findings and the reported association between depressive illness and cardiovascular disease in later life. v

6 Acknowledgements First and foremost, I am grateful to my academic supervisors, Professor Lawrence Beilin, Professor Trevor Mori and Dr. Monique Robinson, for their strong support and advice with regard to all aspects of my studies. My special thanks to Ms. Sally Burrows; her enthusiastic consultancy in statistics has been instrumental in my analyses of data throughout the project. I would like to express my deep gratitude to the families participating in the Raine Study, and acknowledge the support of the Raine Study Executives, the Raine Study team and manager Jenny Mountain. I would like to acknowledge the support from the Commonwealth APA-scholarship Award, University of Western Australia PhD Top-Up scholarship, and the Raine study PhD scholarship. I acknowledge the travel expense support to attend the conference at Adelaide from the high Blood Pressure Research Council of Australia. I also acknowledge the travel expense support to attend the conference at Cape Town, South Africa from UWA and the Raine study. I would like to thank my colleagues in the Raine study and co-authors of my manuscripts who contributed to this thesis. It has been an enjoyable experience to have worked at the Royal Perth Hospital campus of the UWA School of Medicine and Pharmacology over the past four and a half years. Last but not the least my sincere thanks to my family for their support and encouragement and Joanne Edmondston, Graduate Research Officer, UWA for her help in organising my abstract and review of the literature. vi

7 Organisation of thesis This thesis is submitted as a series of papers and comprises six chapters. The first chapter presents the review of the literature which ends with thesis scope and aims. Chapter 2 presents the methodology of the Raine Study cohort. Three papers are presented in chapters 3 to 5. Each paper is preceded by a brief preamble. An overarching discussion is presented in chapter 6 with an interpretation of the findings pertinent to each paper in the context of existing literature, discussing strengths and limitations, drawing overall conclusions and suggesting future directions. Bibliography, containing all references, is placed as Appendix C, at the end of this thesis. Chapter 3 has been published in the Journal of Hypertension and is included as Appendix B in the original published typeset. The articles in chapters 4 and 5 are presented as manuscripts and have been submitted to journals for publication. vii

8 Table of Contents Thesis Declaration... ii Authorship Declaration... iii Abstract... iv Acknowledgements... vi Organisation of thesis... vii Table of Contents... viii List of Figures... x List of Tables... xi List of Abbreviations... xiii Chapter 1 Literature review Cardiovascular Disease (CVD) High BP Adiposity Depression - the non-traditional CVD Risk Factor Review of the evidence on the relationship between depression and obesity Review of the evidence concerning the relationship between depression with BP Other psychosocial/family determinants of CVD CVD risk factors- high BP and obesity- in context of Developmental Origins of Health and Disease (DOHAD) Background Prenatal life stress and DOHAD Review of the evidence on the relationship between maternal distress in pregnancy and offspring BP Gaps in knowledge Scope and aims of the thesis Chapter 2 The Western Australian Pregnancy Cohort (Raine) Study Chapter 3 Contrasting effects of prenatal life stress on BP and BMI in young adults Preamble Abstract Introduction Material and methods Results Discussion References viii

9 Chapter 4 The role of maternal and offspring life style behaviours in the relationship between depression and adiposity in young adults Preamble Abstract Introduction Material and methods Results Discussion References Chapter 5 Relationships between Depression and Anxiety Symptoms Scores and BP in Young Adults Preamble Abstract Introduction Material and methods Results Discussion References Chapter 6 Overall Discussion Key findings in the context of the literature Limitations and strengths Conclusions and future research Appendix A Questionnaires A.1 Prenatal Life Stress Inventory at 18-week Gestation A.1.1 Prenatal Life Stress Inventory at 34-week Gestation A.2 DASS Inventory at Age Appendix B Publication(s) B.1 Contrasting Effects of Prenatal Life Stress on BP and BMI in Young Adults Appendix C Bibliography ix

10 List of Figures Figure 1.1. Depressive symptoms as a risk factor for mortality (adjusted risk estimates using hazard ratios) Figure 1.2. Conceptual model of the relationship between depression and cardiovascular disorders Figure 1.3. Proposed step-by-step model of obesity causation Figure 1.4. Obesity prevalence rates among birth cohorts in famine and control Figure 2.1. The Western Australian Pregnancy Cohort (Raine) Study participants from birth to the 20-year review Figure 3.1. Flow diagram of Raine Study participants attending the 20-year follow-up Figure 3.2. Interaction plot depicting adjusted means of young adult systolic blood pressure for prenatal life stress events score, higher BMI levels accentuating the inverse relationship Figure 4.1. Flow diagram of Raine Study participants attending the 20-year follow-up Figure 4.2. Interaction plot depicting adjusted means of young adult BMI for maternal prenatal smoking levels, maternal prenatal smoking showing a significant positive association between depression score and BMI compared to depression-score slope for offspring of mothers that did not smoke during pregnancy Figure 5.1. Flow diagram of Raine Study participants attending the 20-year follow-up No table of figures entries found. x

11 List of Tables Table 1.1. Dutch time trends in cardiovascular risk factors: main findings of the analyses... 3 Table 1.2. Mean weight change (in kg) over 12 years ( ) according to baseline age: the AusDiab Study... 5 Table 1.3. Description of epidemiological studies of depression and BP Table 1.4. Major DOHAD epidemiological studies (late 1980s to date) Table 1.5. Three contemporary pregnancy cohort studies examining maternal distress during pregnancy and offspring BP Table 1.6. Relationship between prenatal depressive symptoms and BP Table 2.1. Comparison of Raine Study cohort participants at age 20 with contemporaneous Western Australian (WA) Census population (2011 census data) Table 2.2. Comparison of participants and non-participants at 20-year follow-up by infant characteristics at birth Table 3.1. Comparative characteristics of mothers of Raine Study participants versus mothers of non-participants in the 20-year follow-up Table 3.2. Prenatal characteristics of the Raine Study participants mothers Table 3.3. Characteristics of the Raine young adult study participants stratified by gender Table 3.4. Multivariable linear regression outcomes of the association between exposure to prenatal life stress events and BP at 20 years Table 3.5. Standardized regression coefficients (beta coefficient) derivation following multivariable linear regression modelling of the association between exposure to prenatal life stress events and BP at 20 years Table 3.6. Comparative modelling to assess breast feeding, familial BP and maternal DASS score covariates for Y20 SBP and DBP outcomes Table 3.7. Longitudinal mixed effect model analysis for repeated SBP outcome measures Table 3.8. Linear mixed model covariate adjusted estimates for SBP and BMI as a function of prenatal life stress events exposure and their interaction with time Table 3.9. Multivariable logistic regression outcomes of the association between exposure to prenatal life stress events and young adult systolic BP (Prehypertension/Hypertension combined) Table Multivariable logistic regression outcomes of the association between exposure to 3 prenatal life stress events and young adult systolic BP (Prehypertension/Hypertension combined) Table Multivariable linear regression outcomes of the association between exposure to prenatal life stress events and BMI at 20 years Table Comparative modelling to assess breast feeding, maternal DASS and physical activity covariates for Y20 BMI outcome Table Longitudinal mixed effect model analysis for repeated BMI outcome measures Table Multivariable logistic regression outcomes of the association between exposure to prenatal life stress events and young adult Overweight/Obesity combined xi

12 Table 4.1. Pregnancy related characteristics for mothers of participants versus mothers of non-participants in the 20-year follow-up Table 4.2. Participants characteristics by gender Table 4.3. Univariate and Multivariable adjusted BMI regression Table 4.4. Multivariable adjusted BMI regression (Final Model) Table 4.5. Characteristics by maternal smoking or low family income during pregnancy Table 4.6. Multivariable adjusted BMI regression substituting family income at pregnancy for maternal prenatal smoking Table 5.1. Pregnancy related characteristics for mothers of participants compared with mothers of non-participants in the 20-year follow-up Table 5.2. Participants characteristics Table 5.3. Participant characteristics according to presence or absence of depression Table 5.4. Univariate BP regression Table 5.5. Multivariable adjusted systolic BP regression on depression-score (Final Model) Table 5.6. Multivariable adjusted systolic BP regression on anxiety-score Table 5.7. Multivariable adjusted systolic BP regression on self-reported clinical depression Table 5.8. Systolic BP regression on self-reported history of depression and gender interaction No table of figures entries found. xii

13 List of Abbreviations ABCD: Amsterdam Based Children and Development Study ADI: Anxiety and Depression symptom index ADS-Y: Anxious and Depressed scores for Youth BDI-Y: Beck Depression Inventory for Youth BMI: body mass index BP: blood pressure BW: birth weight CBCL: Child Behaviour Checklist CES-D: Center for Epidemiological Studies Depression Scale CHD: coronary heart disease CI: confidence interval CIDI: Composite International Diagnostic Interview CVD: cardiovascular disease DASS: Depression Anxiety Stress Scale DOHAD: developmental origins of health and disease DSM: Diagnostic and Statistical Manual FAD: McMaster Family Assessment Device GAD: Generalized Anxiety Disorder Scale GWB-A: General Well Being questionnaire for Anxiety GWB-D: General Well Being questionnaire for Depression HADS: Hospital Anxiety and Depression Scale HPA axis: hypothalamic pituitary adrenal axis xiii

14 HPA: hypothalamic-pituitary-adrenal axis NHANES: The National Health and Nutrition Examination Survey OC: oral contraceptive OR: odds ratio PHQ: Patient Health Questionnaire PSE: Present State Examination scale PSF: Psychiatric Symptom Frequency scale SBP: systolic blood pressure SD: standard deviation SES: socioeconomic status SMR: standardised mortality rate STAI: Spielberger Trait Anxiety Inventory WHO: World Health Organisation xiv

15 Chapter 1 Literature review 1.1 Cardiovascular Disease (CVD) CVD accounts for nearly 50% of all deaths in the developed world and 25% in the developing world. By 2020, CVD is predicted to claim 25 million deaths each year 1. CVD is the most common cause of death both in the US (American Heart Association, 2013) and worldwide (World Health Organization, 2013) 2. In Australia, approximately 4.2 million adults (18.3%) are reported as having CVD in In 2015, (29%) deaths had an underlying cause related to CVD 3. Population increase, increasing life expectancy, improvements in health screening, and economic, social and cultural changes are the major factors driving the increase in CVD prevalence. A range of biological and psychosocial factors are known to contribute significantly to the pathogenesis and expression of CVD. One of the earliest studies of CVD risk factors was the Framingham Study of 5209 adults (aged 30-62) resident in Framingham, Massachusetts, from 1948 onwards. The study found that raised BP, tobacco use, and raised blood cholesterol, with cumulative effects and diabetes are major risk factors for CVD mortality 4-6. In the Seven Countries study 7 in the late 1960s, a higher dietary ratio of monounsaturated to saturated fatty acids explained to some extent the variance in the study participants CHD rates. This was additional to the variance explained by Framingham CVD risk factors. The list of risk factors now includes lifestyle factors such as medical history of drug treatment for hypertension or diabetes, physical inactivity, unhealthy diet, and alcohol consumption 8. Hypertension, tobacco use, high lipids, diabetes, obesity, alcohol use and low fruit and vegetable intake account for 61% of total CVD death estimates 9. Subsequent studies have confirmed hypertension, tobacco use, high lipids, diabetes, and obesity comprised major epidemiological risk factors for CVD. In the late 1970s, MONICA (MONItoring the trends and determinants of CArdiovascular disease) 10 used repeated cross-sectional surveys and specially created disease registers to assess CVD risk in 38 populations across 21 countries. These surveys profiled those risk factors, including their change, within each geographically defined population. An epidemiological study was also undertaken in Finland in the late 1970s to investigate cardiovascular health and risk factors in children, adolescents and young adults 11. In 1

16 2004 the INTERHEART study 12 examined the effect of potentially modifiable risk factors associated with acute myocardial infarction in 52 countries. The study showed the five risk factors of hypertension, tobacco use, high lipids, diabetes, and obesity attributed approximately 80% of the population attributable risk for CHD 12. These five risk factors found consistently across the range of studies are commonly referred to as the traditional CVD risk factors. A number of studies have assessed risk factors in children that are predictive of CVD in adults. These studies included the Bogalusa Heart study 13,14 ( ), the Muscatine Heart study 15 ( ), and the Cardiovascular Risk in Young Finns study 16,17 ( ). All three studies demonstrated that a considerable number of children had risk factors such as distribution of serum cholesterol and serum triglyceride levels, blood pressure, and body weight which in adults are predictive of coronary heart disease (CHD). Later, the multiethnic Coronary Artery Risk Development in Young Adults (CARDIA) study 18 ( ) found BP reactivity to psychological stress predicted hypertension in young adults 19. The CVD risk for an individual can be alleviated by the modification of lifestyle-related factors and early treatment of CVD risk factors. Preventive programmes initiated by the World Health Organization (WHO) have targeted blood pressure (BP), tobacco use, cholesterol, or all three of these risk factors, which in combination account for 75% of the CVD burden 20. Over the last 25 years these programs have reduced age-standardised CHD mortality by more than 50% in high-income countries, though the trend is flattening or increasing in some low-and-middle income countries 21. In Australia and New Zealand, around 80% of the decline in CHD mortality has been attributed to changes in risk factors particularly diastolic BP, tobacco use, and cholesterol 22. However, the decline in coronary heart disease mortality has begun to slow down, more so in young adults. In Australian men and women aged years, the CHD mortality decline has slowed since the early 1990s and for men aged years, there was no annual percentage change for the period Whereas, a marked deterioration of clinical care in Australia or other high income countries seems implausible, the most plausible explanation is a diminution in decline, arrest or reversal in major CHD risk factors reflecting lifestyle changes in young and middle aged adults 23. The possible reasons are poor diet, a more sedentary lifestyle, and increased overweight/obesity and diabetes. 2

17 1.1.1 High BP Hypertension is the most common cardiovascular condition, accounting for 13% of deaths annually 24 and 4.5% of the disease burden worldwide, contributing more burden than either alcohol or cigarette smoking 25. Of the approximately 17 million deaths that are attributable to CVD globally every year complications of hypertension account for over 55% deaths 26. Hypertension is responsible for at least 45% of deaths due to heart disease and 51% of deaths due to stroke 27. Adults with BP above 130-9/85-9 mmhg are at higher risk of developing clinically defined hypertension or another major cardiac event 28,29. Suboptimal BP control accounts for 62% of cerebrovascular disease and 49% of ischaemic heart disease 25. Every 2 mmhg increase in SBP in adults is associated with a 4% increase in CHD and 6% increase in stroke 30. Successful preventive programmes and improvements in hypertension diagnosis and treatment, are showing a decline in CVD risk factors. The latest set of national trends in traditional risk factors for CHD from the Netherlands, between 1988 and 2012, shows a decreasing trend in BP 8 (Table 1.1). Four out of six risk factors for CHD showed a favourable or stable trend. There was a rise in diabetes mellitus and BMI. The results demonstrated favourable trends in smoking (except in women 65 years) and physical activity (except in men 65 years). Table 1.1. Dutch time trends in cardiovascular risk factors: main findings of the analyses Adults <65 years Elderly 65years Men Women Men Women SBP = = Total cholesterol = = BMI = Smoking = Diabetes Physical activity = Adapted from: Koopman et al (2016) 8 SBP systolic blood pressure, BMI body mass index, increasing time trend., decreasing time trend. =, stable time trend On average, global population SBP decreased slightly since 1980, but trends vary significantly across regions and countries 31. Prevalence of elevated BP in adolescents aged years has flattened in China and the US, despite upward obesity trends in 3

18 the interval 32. Trends in age-standardised mean SBP in Australasia between 1980 and 2008 showed a downward change of 2 3 mm Hg per decade ( 4 5 to 0 3) for men and a downward change of 3 9 mm Hg per decade ( 6 4 to 1 6) for women 31. The ageadjusted prevalence of hypertension among the US adults was 29.1% in and this figure did not change appreciably since The proportion of adult elevated BP attributable to childhood elevated BP is approximately 10%, the degree of tracking from childhood to adulthood somewhat dependent on relative improvements in the factors associated with a healthy lifestyle 34. High BP in childhood is thus partially predictive of high BP in adult life 35. The baseline AusDiab survey 36 ( ), showed one in three Australian adults aged 25 years and over were hypertensive. Hypertension was defined having BP 140/90 mmhg or taking BP-lowering medication. More recent follow-up in showed steeper decline in the incidence of hypertension in age group 25-34y among females (0.9% per year) compared to males (1.3% per year) 36. Those classified with normal BP (<140/90 mmhg and not taking BP medication) at baseline, 27.9% had developed hypertension at 12y follow-up in and those classified with hypertension at baseline, 13.7% were having normal BP at the follow-up Adiposity Adults who are obese have a 1.2 to 1.5 times increased risk of CVD, independent of blood pressure, cholesterol and smoking status 37,38, with obesity defined as abnormal or excessive fat accumulation that may impair health 39. The body mass index (BMI) is a value derived from the mass (weight) and height of an individual, defined as the body mass (in kilograms) divided by the square of the body height (in metres), and is universally expressed in units of kg/m 2. Adults with a BMI between are considered normal weight, between overweight and 30 or above obese. Adiposity is an independent and an important risk factor for hypertension, type 2 diabetes mellitus, hyperlipidaemia. Overweight is a preventable condition that poses a challenge in the prevention of CVD 40. Worldwide, the proportion of adults with a BMI of 25 kg/m 2 or greater increased between from 28 8% (95% CI ) to 36 9% ( ) in men, and from 29 8% ( ) to 38 0% ( ) in women 41. Increasing trends for overweight and obesity are also evident from the UK 42, in the Netherlands(Table 1.1) 4

19 and in many European countries 43. In Western Australia, a study of anthropometric measures taken at 9, 18 and 25 years of age shows overweight or obesity levels are increasing consistent with international trends 44. The AusDiab survey 36 showed weight and waist circumference of Australian people aged years at baseline increased over the 12 years of follow-up in These increases became less with increasing age (Table 1.2). In Australian adults aged year old, mean BMI increased over time from 23.7 (SD 3.5) for males and 22.4 (SD 4.0) for females in 1995 to 25.2 (SD 4.8) and 25.5 (SD 5.9) in At the same time, the prevalence of morbid obesity increased from 0.2% to 1.8%, class 2 severe obesity increased from 0.3% to 2.0%, and class 3 obesity increased from 0.1% to 0.5% among Australians aged 7-15 years 45. Table 1.2. Mean weight change (in kg) over 12 years ( ) according to baseline age: the AusDiab Study Age group (age in years) Men (kg) Women (kg) Adapted from: Tanamas et al (2013) 36 Obesity is an important risk factor for hypertension 46,47. A high body mass index (BMI) is associated with higher BP levels in adults and for every 1 kg/m 2 higher adult BMI, systolic BP (SBP) increases on average by approximately 1 mmhg 48. Separate studies of 1946 and 1958 British birth cohorts show high BMI and large BMI gain during adulthood to be associated with higher adult BP levels 49,50. Given high BP in childhood is predictive of high BP in adult life 35, preventive interventions that aim to reduce the distribution of BP and BMI in children can deliver a substantial reduction in CVD incidence in adults. 5

20 1.2 Depression - a non-traditional CVD Risk Factor Depression is a negative emotional state frequently present in patients with CHD. Major depression or the current diagnosis of major depressive disorder relies on clinical features defined in the Diagnostic and Statistical Manual of Mental Disorder (DSM- V) 51. It is a relatively new risk factor for CVD deaths that explains additional variance, beyond the traditional lifestyle risk factors, in CHD rates 52. Depression is associated both with increased CHD risk and CHD mortality, but is yet to be listed as an official CVD risk factor in the guidelines 53,54. Results from a review of eleven cohort studies, with clinical depression or depressive mood as the exposure and myocardial infarction as the outcome, show that clinical depression (RR=2.69, 95% CI= , p<0.001) is a stronger predictor for the development of CHD in initially healthy people than depressive mood (RR=1.49, 95% CI= , p=0.02) 55. However, clinical trials that have examined whether treating depression decreases the risk of cardiac events in patients with established CHD show that treatment results do not improve cardiac outcomes 56. Several biological and behavioural mechanisms have been hypothesized to underlie the relationship between depression and CHD, but none has been shown to account for more than a small proportion of the risk 56. The mechanisms that may be responsible for depressive symptoms affecting the cardiovascular system are not known. Additionally, it is unclear whether treatment of depression would prevent or reduce CVD. Depression has yet to be established as an independent risk factor for CHD possibly because of incomplete and biased availability of adjustment for conventional risk factors and severity of coronary disease 57. Besides insufficient evidence to demonstrate that depression treatment improve cardiac outcomes 56 there is some evidence of causality among plausible underlying mechanisms that link depression to cardiovascular disease. For instance, immuno-inflammatory and metabolic dysregulations are especially pronounced in atypical depression patients and in cardiovascular disease 58. A meta-analysis of 20 studies published in 29 papers, males (n=8573) and females (n=2422), assessing the role of depression as a risk factor for mortality in CHD patients found depression had an unfavourable impact on mortality in CHD patients 52 (Figure 1.1). 6

21 Figure 1.1. Depressive symptoms as a risk factor for mortality (adjusted risk estimates using hazard ratios). Adapted from: Bath et al (2004) 52 Depression, once believed to be a mental disorder restricted to adults, is now known to affect children as well. Children who have depression are more likely to be diagnosed comorbid with anxiety disorders 59. a) Depression burden and relationship with CVD Depression and anxiety disorders (negative affect) are integral components of a broad group of Mood disorders, the group of diagnoses in the Diagnostic and Statistical Manual of Mental Disorders (DSM) classification system 60. In adults a 12-month prevalence of Mood disorders, including depression disorders, varied from as low as 3% in Japan 61 to over 9% in the US 61. A lifetime prevalence has a range from 6-15% across surveys 41,61, some suggesting one or more episodes of depression during the life time among half of the population 62. More than 1 out of 20 people in the US twelve years of age and older reported current depression (moderate or severe depressive symptoms in the past 2 weeks) in Depression in clinical settings is more researched than anxiety disorders in part because of the paucity of brief validated measures for anxiety compared with the numerous measures for depression 64,65. One of the most common anxiety disorders seen in general medical practice and in the general population is generalized anxiety disorder 66. The disorder has an estimated current prevalence in general medical practice of 2.8% to 8.5% 67,68. DSM-IV requirement that episodes of generalized anxiety disorder must persist for at least 6 months may exclude a large number of people that suffer from anxiety symptoms with episodes of less than 6 months duration 69. 7

22 The Dunedin Study in New Zealand is the longest ongoing longitudinal investigation of health and behaviour of a complete birth cohort with study participants (n=1037) born between April 1st 1972 and March 30 th, In the Dunedin study cohort, 16.8% had depression disorder at 21 years 70. Globally, major depressive disorder increased from 15 th to 11 th rank (37% increase) in disability-adjusted life years (calculated as the sum of years of life lost and years lived with disability) from 1990 to The increase in the negative affect associated with depression and anxiety over the past few decades has been phenomenal. The Dunedin Study 70,72 reported that the prevalence of anxiety and depression was about twice as high as estimated and provided evidence that lifetime prevalence rates are doubled by prospective versus retrospective ascertainment. Major depression is a strong risk factor for the development of CVD, estimated by some to equal the burden of risk associated with smoking or diabetes 73. Data from the Netherlands suggests that patients with major depression are more likely to show features of the metabolic syndrome that includes a high BMI and an increased risk for type 2 diabetes, CHD, and hypertension compared to those with a moderate severity of depressive illness 74. The relationship between depression and CVD is complex. Some of the mediators that correlate with depression and CHD are poor health behaviour, maladaptive coping style, social isolation, chronic life stress, smoking, low physical activity, a poor diet, and noncompliance of medical recommendations 52,75,76. Figure 1.2 illustrates how patients with depression and cardiovascular disorders suffer from a mutually reinforcing cycle that can worsen both mental and physical health 77. 8

23 Figure 1.2. Conceptual model of the relationship between depression and cardiovascular disorders. Adapted from: Whooley et al (2013) 77 b) Measurement of depression symptoms Epidemiological studies have largely used questionnaires to ascertain depression or anxiety symptom scores but in some instances, have relied on various forms of clinical assessment of major depression to categorize people for severity. Some questionnaires have used categories to define significant depression rather than using continuous scales. The following twelve validated inventories are commonly used to ascertain depression and/or anxiety scores, namely: CES-D (The Center for Epidemiological Studies- Depression scale); PHQ (Patient Health Questionnaire); GAD (Generalized Anxiety Disorder Scale); HADS-D (Hospital Anxiety and Depression Scale); STAI (Spielberger Trait Anxiety Inventory); CIDI (Composite International Diagnostic Interview); GWB (General Well Being questionnaire for depression or anxiety); PSE (Present State Examination scale); PSF (Psychiatric Symptom Frequency scale); BDI (Beck Depression Inventory); ADI (Anxiety and Depression symptom index); and the Child Behaviour Checklist (CBCL). A commonly used inventory in Australian epidemiologic research is the Depression Anxiety Stress Scale (DASS-42) 78. The DASS-21 is a short form of the 42-item DASS 78 9

24 with both scales found to have good reliability and validity in both clinical and nonclinical samples The DASS-21 is a 21-item self-report measure comprised of three sets of seven items measuring depression, anxiety and stress respectively. The seven symptom areas of the DASS Depression scale (Dysphoria, Hopelessness, Devaluation of life, Self-Depreciation, Lack of interest/involvement, Anhedonia, Inertia) are not only well represented in the other self-report scales as any seven other symptoms but are reasonably well-represented in the depression symptom lists of other authoritative sources 78,82. The depression scale score measures similar criteria to those used to diagnose Diagnostic and Statistical Manual (DSM) Mood Disorders 83. DSM- IV is a categorical classification that divides mental disorders into types based on criterion sets with defining features. However, a dimensional model of depression and anxiety symptoms is favored over the diagnostic categories of mood disorders and anxiety disorders 84. The anxiety scale measures autonomic arousal, skeletal musculature effects, situational anxiety and subjective experience of anxious affect. The stress scale assesses difficulty relaxing, nervous arousal, being easily upset/agitated, irritable/overreactive and impatient 78. Participants are asked to rate the severity of each symptom during the past week on a four-point scale ranging from zero ( did not apply to me at all ) to 3 ( applied to me very much, or most of the time ) and scores doubled. Several studies have suggested temporal stability of the DASS across time A large scale study has shown stability of symptoms as measured by the DASS over 3 to 8 years 89. Analyses specific to the DASS-21 have shown a quadripartite structure, which consisted of a general factor the authors suggested reflected general psychological distress, and independent factors suggested to represent depression, anxiety and stress

25 1.2.1 Review of the evidence on the relationship between depression and obesity Depressed patients are more likely to be overtly obese than healthy controls 90 and are more likely to become obese at follow-up than non-depressed people 91. The National Health and Nutrition Examination Survey (NHANES) from the US shows 43% of adults with prevalent depression symptoms were obese 63. Obesity is associated with an increased risk of mood disorders in both genders 92, but the odds for increased risk of depression are higher with extreme obesity (BMI 40kg/m 2 ) 93. The increasing BMI trends over the decades 94, and an increasing incidence of overweight and obese children and adolescents 95 is not only paralleled by an increase in mental health disorders in adolescents 96, but is invariably accompanied by the co-occurrence of depression and obesity in young adults 63,97,98. There is complexity in the relationship between depression and obesity. For instance, socioeconomic disadvantage may cause psychological and emotional distress, or a disharmonious family environment, or psychological and emotional overload, or disruption of energy homeostasis leading to obesity (Figure 1.3). Hemmingsson 99 has, proposed a theoretical model that posits psychological and emotional distress as a fundamental link between socioeconomic disadvantage and weight gain. The socioeconomic and psychological factors affect the behavioural lifestyles such as overeating, poor food choices, or lack of physical activity. Figure 1.3. Proposed step-by-step model of obesity causation. Adapted from: Hemmingsson (2014) 99 11

26 Psychosocial factors such as financial difficulties and familial factors have a cumulative effect on CVD 12,100 and a possible aetiological role for CVD 101. Increases in income inequality, which place children into low SES conditions, renders a negative effect on behavioural and psychosocial health. Adult behaviour and psychosocial status is patterned by childhood SES 102. Early socioeconomic disadvantage influences later BP in part through an effect on BP in early life, and in part through an effect on BMI 103,104. Social position is consistently linked with obesity. The strength of the association between social position and obesity is consistently stronger than either diet or physical activity, and appears to be growing in strength There is a strong evidence of obesity linked to socioeconomic disadvantage in children and adults Obese women with low income, or no more than a high school education are more likely to experience moderate-severe depressive symptoms than their respective counterparts who are older and more affluent 110. SES not only is inversely associated with BMI and contributes to the co-occurrence of depression and obesity 91, , but accounts for 79% of the association between major depression and BMI 115. Psychosocial distress, resulting from harsh parenting, acts as a mediatory pathway in the relationship between childhood SES and poor metabolic functioning 116. Childhood neglect is also associated with depression and obesity 117. Other than obesity, children from lower SES families also have a propensity for higher BP 118. a) Gender differences There are gender differences in the relationship between obesity and depression. Goldney et al 119 found that obese Australian men were less likely to have major depression compared to those of normal weight (OR = 0.46; 95%CI: 0.24, 88) whereas overweight women reported less psychological distress than those of a healthy weight. A national survey from the US found obesity in males was associated with a 37% decrease in major depression, whereas obesity in females was associated with a 37% increase in major depression 120. The Zurich cohort study showed that among men, depressive symptoms before the age 17 are associated with increased weight gain (6.6 vs 5.2% BMI increase per 10 years) in adulthood but not obesity 121, whereas among women, depressive symptoms before age 17 are associated with increased weight gain (4.8 vs 2.6% BMI increase per 10 years) and greater risk of adult obesity (hazard ratio=11.52, P<0.05) 121. Kasen et al 122 found young female adults with baseline BMI 30kg/m 2 were at a significantly increased risk for subsequent major depression and generalized anxiety disorder. NHANES also showed obesity was related to 12

27 depression severity only among women (age-adjusted); one-half of women with severe depressive symptoms were obese compared with one-third of women with no depressive symptoms 63. The mental health of women is more closely affected by overweight and obesity than that of men 97,98 and the perception of being obese appears to be more predictive of mental disorders than actual obesity 123,124. Depressed female adolescents are more likely to become obese as adults 97. b) Possible psychosocial factors underlying the association between depression and obesity Offspring obesity could in part be due to childhood lifestyle factors such as dietary habits and physical inactivity 125, and in part to the influence of adverse maternal behaviours that includes maternal prenatal smoking. Li et al 126, in a cross-sectional study of Portuguese boys and girls (aged 3-10y), also found children of mothers that smoked during pregnancy had a higher adiposity level than children of the mothers that did not smoke, with a median difference in BMI of 0.39 kg m 2 (95%CI: 0.25, 0.53) in boys and 0.46 kg m 2 (95%CI: 0.31, 0.62) in girls. The association between maternal prenatal smoking and incident hypertension reflects a lifetime accumulation of postnatal psychosocial influences 127, with continuance of smoking during pregnancy being strongly correlated with low maternal education 128. Smoking in pregnancy constitutes an area of women's behaviour that is linked systematically with aspects of their material and social position 129.Put another way, smoking in pregnancy is associated with conditions of material disadvantage, with social stress, anxiety and indices of low control over living conditions; and with low levels of social support 129. Parental and early childhood influences including maternal education and smoking in pregnancy are important early influences on adolescent obesity 130. Maternal prenatal smoking is a possible marker of other socioeconomic determinants and behaviours. An association between childhood SES and adult obesity has been reported 131. Also, there is a strong association between socioeconomic factors and depression in both men and women 132. Maternal smoking during pregnancy has been associated with reduced duration of breastfeeding and reduced time spent breastfeeding is associated with a higher BMI 133. There is strong evidence of social patterning with higher rates of negative affect for young people from poorer homes, with lower social class 134, and with younger, less educated mothers 135. Such disadvantaged families create vulnerabilities and promote a poor bio-behavioural profiles of their offspring that lead to consequent accumulating risk for health disorders 136. In contrast, family resilience 13

28 involves sharing beliefs and emotions within a family that ultimately leads to conflict resolution and reduced chronic strain 137,138. Childhood family environments represent vital links for understanding mental and physical health across the life span 136,139. In brief, smoking in pregnancy is associated with conditions of material disadvantage 128,129, has shared variance with other family psychosocial covariates of young offspring, and is an important early influences on offspring obesity 130,140,141 and depressive/anxiety behaviours Therefore, any relationship between offspring depression and BMI may be amenable to modification by maternal prenatal smoking. The major studies 119,121,122 that examined the relationship between depression and adiposity did not examine the potential interaction effects of prenatal smoking on the relationship between offspring depression and adiposity. Over the past decade there is parallel increase in childhood mental health disorders 96, and adiposity among children and adolescents 95. Also, the two conditions often co-exist 98,145. Findings from the NHANES (survey ) also show 7.2% of the U.S. adults aged 20 and over had current depression, 34.6% of young adults were obese, and 43% had co-existing depression and obesity 63. Given a strong evidence of social patterning with higher rates of negative affect for young people 134,135, an assessment of interaction effect of prenatal smoking on the relationship between offspring depression and adiposity relationship may disentangle the causal and psychosocial influence of maternal prenatal smoking on the association between offspring depression and adiposity. Chapter 4 of this thesis explores the interaction effects of maternal smoking in pregnancy on the relationship between offspring depression and BMI, at age 20. The question of interest is how would maternal smoking in pregnancy influence the relationship between offspring depression and BMI. c) Possible mechanisms underlying the association between depression and obesity A potent physiological mechanism that mediates the relationship between depression and obesity involves activation of the hypothalamic pituitary adrenal (HPA) axis 146. Heightened levels of cortisol contribute to weight gain, with a particular impact on increased visceral adiposity 147. Inflammation is the other possible physiological link between depression and obesity, as production of high levels of pro-inflammatory cytokines have been linked with both depression 148 and obesity 149. Miller et al 149 found depressed subjects had greater body mass than control subjects, but adiposity only accounted for part of the link between clinical depression and increased expression of 14

29 inflammatory markers. Psychosocial stress can elevate plasma levels of interleukin and stimulate the HPA axis, increasing propensity for obesity and CVD. Few studies have found high plasma levels of leptin in both depression and obesity, implicating leptin resistance for those high plasma levels 151,152. Adiposity accompanied by psychological and emotional overload affects physiological energy homeostasis systems and delivers an increased level of cortisol, ghrelin, insulin and pro-inflammatory cytokines, in response to an increased HPA axis dysregulation Any disruption of energy homeostasis culminates in weight gain mediated through behavioural disturbances including emotional eating 153, In obese children, reduced sympathetic as well as parasympathetic nerve activities could cause energy homeostasis dysfunction Review of the evidence concerning the relationship between depression with BP Table 1.3 lists some of the major studies of the 21 st century that examined the relationship between depression and BP. Specific criteria applied for searching the articles relating to population studies in PubMed were the use of key words or expressions such as depression, blood pressure, or association between depression and blood pressure, hypertension, or hypotension. This literature review considered those methodological studies that focused specifically on depression and blood pressure outcomes. Wherever relevant, the cut-off values used for hypertension are placed along with the list of adjusted confounders to emphasise disparity between the studies. The studies from the US have mostly used the CES-D Scale questionnaire to ascertain depression score, whereas others used depression and/or anxiety inventories pertaining to their geographic location and availability. a) Positive relationships between depression and BP Many studies have found a positive correlation between depression scores and hypertension. Jonas et al 163 showed higher levels of depression and anxiety symptoms (combining two four-item scales from the General Well-Being Schedule) in 3310 participants (initially normotensive and chronic disease-free, aged 25-64y) were associated with higher levels of treated hypertension. Hypertension was defined as SBP 160 mmhg, DBP 95 mmhg, or reporting the prescription of antihypertensive medications. In another study, middle-aged men (n=1038, aged 42, 48, 54, or 60) who reported high levels of hopelessness were found three times more likely to become 15

30 hypertensive. Hypertension was defined as SBP 165 mmhg and/or DBP 95 mmhg or confirmed use of antihypertensive medication. Adjustments were made for age, BMI, baseline resting BP, physical activity, smoking, alcohol consumption, education, parental history of hypertension, and self-reported depressive symptoms 164. In the multiethnic CARDIA study, young adults (n=3343, aged 23-35y at baseline) who had high scores (CES-D 16) were at a significantly greater risk for hypertension. Hypertension was defined as SBP>160 mmhg or DBP>95 mmhg. Adjustments were made for age, resting SBP at the 5 th year examination, physical activity, alcohol use, parental history of hypertension, education, presence of chronic disease, sex, and race 165. Delaney et al 166 found patients with baseline depressive symptoms (aged 45-84y) did not have an increased incidence of hypertension, but patients using tricyclic antidepressants had a positive relationship with hypertension (RR=1.20; 95%CI: 1.05, 1.37). Depression, defined as a CES-D score 16 or the use of any antidepressant medication, was associated with small changes in SBP (+2.4mmHg; 95%CI: 0.2, 4.7) and DBP (+0.8mmHg; 95%CI: -0.6, 2.3). Hypertension was defined as SBP 140 mmhg, DBP 90 mmhg or new use of antihypertensive medications plus physician diagnosis. Adjustments were made for age, sex, baseline BP, diabetes, and BMI. Meng et al 167 conducted a meta-analyses of nine prospective studies (with total participants=22367 and a mean follow-up period of 9.6 years) showing depression increased the risk of hypertension, the risk correlating with the length of follow-up and the prevalence of depression at baseline. In the Australian Longitudinal Study on Women s Health, Jackson et al 168 found among women hypertension-free at baseline (n=12338, mean age 49.5) that depression (using CES-D) was associated with a 30% increased odds of hypertension (age-adjusted OR=1.30, 95%CI: 1.19, 1.43) and anxiety (self-reporting of diagnosed ever with or treated ever ) was associated with a 24% increased odds of hypertension (age-adjusted OR= 1.24, 95%CI: 1.09, 1.42), during 15- year follow-up from 1998 to 2013 for women born in Adjustments were made for smoking, BMI, physical activity, alcohol intake, SES, marriage relationship status, area of residence, diagnosed with or treated for diabetes mellitus, heart disease or stroke, menopausal status. This study demonstrated the significance of adjustments in statistical analyses. The association between depression and hypertension was markedly attenuated after adjusting for BMI alone (OR= 1.19, 95% CI: 1.08, 1.31) suggesting that the relationship was to a considerable extent explained by adiposity. The association 16

31 between anxiety and hypertension was no longer significant after they adjusted for age and depression (OR=1.15, 95%CI: 0.99, 1.32). This finding supports adiposity as a strong confounder for an association between depression and hypertension. What can be considered as a limitation of this study is the failure to ascertain if the association between depression and hypertension was similar at various BMI levels. For instance, the influence of obesity on the association may be different to the influence of normal or overweight BMI on the association between depression and hypertension. Thus, adiposity interaction on any association between depression and hypertension may reflect more meaningful estimates than simple adjustment for adiposity in a model. b) Inverse relationships between depression and BP In contrast to the above reports Hildrum et al 169 found in Norwegian men and women (aged 20-67y) participating in the Nord-Trøndelag Health Study (HUNT) an inverse association between depression and/or anxiety symptoms score and mean SBP (-1.59 mmhg, P=0.004) and DBP (-0.78 mmhg, P=0.019) and a 20% (P=0.001) lower risk of developing hypertension at year 22 (BP 140/90 mmhg). They used the Hospital Anxiety and Depression Scale (HADS), and Anxiety and Depression symptom index (ADI) inventories to examine the association between depression/anxiety scores and development of hypertension, using repeated assessments of anxiety, depression and BP. Among a wide range of age groups, the study showed that the inverse associations did not vary significantly within sex or age group stratified results. Louise et al 170 found in 706 boys and 680 girls of the Western Australian Pregnancy Cohort (Raine) study participants (aged 14y) that boys with higher anxious-depressed scores (using CBCL) had a lower BP. In the longitudinal analyses, boys with higher anxious-depressed scores also had lower SBP trajectories. After adjustment, the predicted mean difference between SBP of boys with a depressive symptom score (BDI-Y 15) in the top 5% and those who had no symptoms was 3.0 mm Hg. Adjustments were made for age, age of first menstruation as an indicator of puberty (available for females only), SES, BMI, and lifestyle factors including diet and physical activity. The study found an inverse relationship between depression and anxiety scores and BP with gender differences. Tikhonoff et al 171 examined the relationship between depression or anxiety symptoms (at ages 36, 43, 53 and years and BP (at age 60 64y). The negative affect of depression and anxiety was ascertained using the Present State Examination scale and 17

32 Psychiatric Symptom Frequency scale, respectively. The authors found a cumulative effect of depression/anxiety symptoms lowered SBP. Adjustments were made for sex, BMI, educational attainment, SES, heart rate, lifestyle factors and antihypertensive treatment. Compared with those never meeting the case criteria, participants with symptoms at one to two time-points had a lower SBP (-1.83 mmhg; 95%CI: to 0.01) but with symptoms at three to four time-points SBP was much lower (-3.93 mmhg; 95% CI: to -0.68). This study suggested that the inverse impact of affective symptoms on SBP across adulthood was cumulative. Table 1.3. Description of epidemiological studies of depression and BP Author (et al) Date Published Population(n) Location and design Outcome Positive effects Jonas Everson United States (US); Follow-up study 1038 Finland; Follow-up study negative affect = risk for selfreported, treated, and incident hypertension. White RR=1.73 (95%CI: 1.30, 2.30), black RR 3.12 (95%CI: 1.24, 7.88), and all men RR= 1.56 (95% CI: 1.08, 2.25) High levels of hopelessness OR=3.22 (95%CI: 1.56, 6.67). Moderate [OR=1.27 (0.8,2.1)] for hypertension Davidson Meng Winkel Jani Maatouk Jackson Negative effects United States; Follow-up study Meta-analyses; Prospective studies Germany; Prospective cohort study Scotland: Prospective cohort study Germany; Prospective cohort study Australia; Follow-up study CES-D scores 16 hypertension OR=2.10(1.22, 3.61) Adjusted relative risk 1.42, 95%CI: 1.09, 1.86, P=0.009 Comorbid anxiety/depression had a significantly higher BP (systolic = 3.0, 95% CI = ; diastolic = 2.3, 95% CI = ) Adjusted HR (SBP)=1.28 ( ) Adjusted HR (HADS- D at baseline) =1.22 ( Clinically significant depression & anxiety showed odds of being hypertensive OR=1.76 (1.14, 2.74). and OR=1.1 (0.85, 1.44), respectively. Fully adjusted depression analyses OR= 1.07 (95% CI: 0.96, 1.20) Fully adjusted anxiety analyses OR=1.06 (95%CI: 0.90, 1.24) Stroup- Benham Unites States; Population based study Hypotensive subjects scored a CES- D mean ±.67 vs ±.95 for normotensives Licht Netherlands; Cross-sectional study 18 Participants having remitted or current depression had a lower mean SBP (β= 1.74, P=0.04 and β = 2.35, P=0.004, respectively)

33 Author (et al) Date Published Population(n) Location and design Outcome Hildrum Louise Norway; Prospective cohort study Australia; Pregnancy cohort study 80th percentile of anxiety plus depression = lower mean SBP (-0.67 mm Hg, p=0.044) and DBP (-0.25 mm Hg, p = 0.201) BP at year 22 SBP for depressed boys at 14y: β= (95%CI: , ) Tikhonoff England, Scotland and Wales; National Survey In fully adjusted models, found borderline association at 43y, between anxiety and depression and SBP=-2.71mmHg (-5.47, 0.06) Mixed or no results Shinn United States; Follow-up study OR=0.64 (95%CI: 0.14, 1.45) for depression and OR=0.70 (0.30, 1.64) for anxiety Yan Grimsrud Licht United States; Follow-up study South Africa; Nationally Survey Netherlands; Cross-sectional study OR=1.29 (0.96, 1.74); with other hypertension-cut-off OR=1.49 (1.09, 2.02)] for intermediate depression & 1.51(1.07, 2.13)] for high depression. Hypertension diagnosis was associated with 12-month anxiety disorders OR = 1.55 (95% CI: 1.10, 2.18) but not 12-month depressive disorders Tricyclic antidepressant users had Hypertension stage1 OR= 1.90 (0.94, 3.84) Hypertension stage2 OR= 3.19 (1.35, 7.59) Delaney United States; Prospective cohort study Relative risk of patients with baseline depressive symptoms RR=1.02 (95%CI: 0.99 to 1.05) c) Comparison between studies Apart from the studies by Hildrum et al 169 and Tikhonoff et al 171 that collected repeated measurements over time, other prospective studies collected information on exposures (depression and/or anxiety) and confounders at baseline. These studies indicate indirectly the great difficulty in doing studies of depression and/or anxiety in relation to BP. The general limitations relate to the lack of adequate information on confounders. This is best exemplified by the study findings of Licht et al 176 and Delaney et al 166 that showed a history of medication has a positive influence on the relationship between depression and BP. Patients using tricyclic antidepressants had a significantly higher SBP and DBP compared with healthy controls and non-medicated patients and a higher risk of being clinically hypertensive stage-1 or stage-2. Hypertensive stage-1 equated to SBP 140 mmhg and DBP 90mmHg whereas hypertensive stage-2 equated to 19

34 SBP 160 mmhg and DBP 100mmHg. Users of selective serotonin reuptake inhibitors (WHO Anatomic Therapeutic Chemical (ATC) code N06AB), and noradrenergic and serotonergic working antidepressants (ATC code N06AX) were more likely to have hypertension stage-1. Another difficulty in doing studies of depression and/or anxiety in relation to BP is an earlier time-period setting when clinical care is different, and outcomes vary as per the definition of hypertension cut-off values. For instance, there have been changes in the definition of prehypertension or hypertension in adults over the past decades and selective serotonin reuptake inhibitors or noradrenergic and serotonergic working antidepressants are now preferred over tricyclic antidepressants. Perhaps not having a similar set of variables for confounding adjustment were at play when Jonas et al 163 and Davidson et al 165 found an approximately twofold increase in the odds of hypertension with baseline depressive symptoms among young adults, whereas Shinn et al 177 and Delaney et al 166 did not find any significant relationship between depression symptoms and incident hypertension. Yet another difficulty encountered comparing different studies is regarding the methodological protocols that may vary across the studies. For instance, Tikinoff et al 171, suggested the difficulty of assessing independent effects of anxiety and depression on BP when the assessments ascertaining anxiety and depression are done differently (varying inventories and personnel) at different time-points. Meta-analysing the studies that examined the relationship between depression and BP, Meng et al 167 found that the length of followup, sample size, age, sex and ethnic composition, psychological measures, definition of hypertension, potential mediators and confounders varied across the studies. The risk estimation of depression and hypertension association will vary as per the stage of hypertension chosen for the response. For instance, Yan et al 178 found no significant relationship between depression and incident hypertension, defined hypertension as SBP 140mmHg or DBP 90mmHg, whereas substituting the alternate hypertension definition of SBP >160mmHg, DBP >95mmHg, elicited a significantly positive association. Difference in behavioural lifestyle due to different cultures or mean age difference among studies also plays a significant role in the hypertension outcome. For instance, Grimsrud et al s 179 study of a South African population that showed no significant relationship between clinically diagnosed mood disorder and hypertension, may have been influenced by the difference in behavioural lifestyle of South African population 20

35 versus the Western populations. The South African sample was nationally representative of individuals age 18 years and above, with the sample obtained from diverse housing categories. In another multi-ethnic elderly population, Delaney et al 166 found no relationship between higher depression symptoms and incident hypertension. The elderly age of participants can contribute to an absent relationship between depression and BP, due to a lower HPA axis activity observed in old depressed patients 171, Lower cortisol levels in elderly depressed patients indicate a sign of exhaustion of the HPA axis, especially after chronic and recurrent depressive episodes 171, In contrast, Stroup-Benham et al 175 found in 2723 elderly Mexican Americans (aged 65 or older) and not living in institutions, the existence of a relationship between low BP and higher levels of depressive symptomatology (using CES-D Scale). Whether geographical differences or confounding by inadequate history of medications attributed to the difference in outcomes was not clear. Licht et al 176 found in 2981 adults (aged years) that those having remitted or current depression (using CIDI, v-2.1) had a lower mean SBP (β= 1.74, P=0.04 and β = 2.35, P=0.004, respectively), compared to controls. They showed that depressive disorder is associated with low SBP and less hypertension, whereas the use of tricyclic antidepressants is associated with both high diastolic and systolic blood pressures and hypertension. Occasionally, there is a possibility of reverse causation that should be borne in mind concerning any positive relationship between hypertension and depression/anxiety. The increased symptoms of depression and/or anxiety may be the reactions to a diagnosis of hypertension 183. ESTHER-cohort study 174 elderly participants had a positive association between depression and hypertension, whereas no association was found between symptoms of generalized anxiety and hypertension, despite a higher prevalence of generalized anxiety (13.9%) compared to clinically significant depression prevalence of 5.2%. There is a shared variance between depression and anxiety symptoms but the symptoms of anxiety weigh less compared to the clinical severity of depression on the higher end of the depression-anxiety depression spectrum. The Australian AusDiab study in reported that the prevalence of depression is 8% higher in people with hypertension compared to people with normal BP 36. In the absence of explanation for a higher proportion of depression among hypertensives, reverse causation can be assumed or their age-unstandardized data could be reflecting higher prevalence due to under 21

36 representation of younger age-groups or over-representation of older ages. Jackson et al 168 demonstrated the importance of confounder adjustment. An inadequate adjustment had an impact on their results, a significant and positive depressionhypertension association parameter estimate (unadjusted) was markedly attenuated after adjustment for BMI. Study findings of Jani et al 173 reiterated the importance of interactions in the relation between depression and CVD end point. A common shortcoming of all the above studies with BP outcome was that the studies lacked assessment whether the estimates for relative chance of hypertension differed substantially in the different levels of adiposity for males and females, or differed between those with depression vs. non-depression for obese/non-obese subjects. Examining the interaction between depression and adiposity for hypertension outcome is important. For example, women of all age groups with depression are more likely to be obese than women without depression 63. Obesity is otherwise an independent and a major controllable contributor to the development of hypertension in adults 48,184. Questionnaires used to ascertain depression and/or anxiety status suffer from shortcomings. For instance, the absence of questioning for somatic symptoms or sleep disturbance in the HUNT population studies that demonstrated the inverse relationship between BP and depression or anxiety symptoms 169,185,186 has been criticised 174. Albeit, Tikhonoff et al 171 and Louise et al 170 in the Raine pregnancy cohort found a similar direction for the relationship (inverse) between depression symptoms and BP despite using different questionnaires and with study participants different in ages- middle-age adulthood and early adolescence- respectively. Not many studies have shown gender differences in the relationship between depression and BP in younger populations. In the Western Australian Pregnancy Cohort (Raine) study, a low SBP at age 14 was found to be associated with higher depressive symptom scores among boys but a non-significant relationship among girls 170. In this study, the sexes were analysed separately rather than the statistically more powerful approach of combining them and seeking evidence for a significant gender interaction. In the past, the gender gap in depression has shown widening in adulthood as women and men enter and live out their unequal adult statuses 187. In the Baltimore Longitudinal Study of Ageing in US, a higher SBP was found to be associated with higher depressive symptom scores in women, and a higher SBP was found to be associated with lower depressive symptom scores in men 188. Females show greater increases in depressed 22

37 mood and social disconnection in response to an inflammatory challenge 189, indicating that inflammatory processes may underlie depression. Any relationship between depression and BP ought to benefit more by adjusting for gender interaction effect on the association rather than mere adjustment for gender confounding effect. Because obesity is an independent and a major controllable contributor to hypertension development in adults 48,184 and women with depression are more likely to be obese than women without depression, in every age group 63. Various population based epidemiological cohorts, representing wide range of ages, have consistently shown an inverse association between depression and BP. They were the nationally representative British old age population study 171 and three Hunt studies 169,185,186. Another pregnancy cohort study (Raine Study) showed a gender specific inverse association among adolescent boys only at age Given the consistency of the inverse relationships between depression and BP ranging from adolescence to old age, it is worthwhile to re-examine the relationship among Raine study offspring at young adulthood. Emerging adulthood compared to older ages has less comorbidities. In the background of biologically plausible mechanisms for a positive relationship between depression and hypertension , increasing depression and obesity prevalence, obesity per se being an independent risk factor for high BP, it is more straightforward to accept the rejection of an inverse relationship finding between depression and BP as the study hypothesis. A 20-year-old is not only past the reverberations of adolescence hormonal fluctuations, especially among girls, but relatively free of co-morbidities, possibly, consuming less prescription medications compared to older adults. The Raine Study considers young adult population with well documented confounders. Hence, chances of confounding are relatively minimal in young adults when compared to middle aged adults or the elderly Other psychosocial/family determinants of CVD Psychosocial factors such as financial difficulties and familial factors in private life including social isolation have a cumulative effect on CVD 12,100 and a possible aetiological role for CVD 101. The Cardiovascular Risk in Young Finns study 16,17 was the earliest study that found psychological, behavioural and socioeconomic risk factors inside families were highly accumulated and affected CHD outcomes. Suggesting SES as a risk factor for CVD simply indicates a lack of knowledge about the behavioural, 23

38 social, psychological, and biologic pathways by which SES affects CVD 193. This is further elaborated below. a) Socioeconomic status/education level Rates of mortality and morbidity for almost every disease and condition differ according to SES 194. Yet traditional risk factors in large longitudinal study of civil servants in London (Whitehall study) left unexplained a large part of the subsequent intergrade (between grade of employment) differences in death from CHD 195. Although the prevalence of traditional risk factors such as smoking, high blood pressure, physical inactivity, obesity, and other risk factors were higher in the lower employment grades, these differences did not explain the increased risk of dying from CHD. For instance, men in the lowest grade had 2.7 times the 10-year risk of coronary heart disease death of those in the highest grade. When a multiple logistic model was used to adjust for smoking, systolic blood pressure, plasma cholesterol level, blood glucose level, and height, the relative risk was reduced by less than a fourth (to 2.1). An inverse relation between measures of SES and cardiovascular events may in part be attributed to underlying behavioural, social, psychological, and biologic pathways by which SES affects CVD 193. Moreover, there are no trials in which people are randomly allocated to a particular SES group that would have confided in the direct association between SES and cardiovascular events 193. Observational studies have simply examined the net association between SES and a cardiovascular outcome after statistical adjustment for a variety of accepted (traditional) risk factors (principally considered as analysis of confounding) and any amount of measurement error in the confounder or exposure variable can influence such an association 193. In other words, the association between cardiovascular disease and the SES measure of interest may be confounded by some other risk factor. Differences in lifestyles and behaviour patterns is an important possible pathway by which SES affects cardiovascular disease 196. A high job strain may adversely affect health status despite having a higher socioeconomic status. Low-SES environments foster negative health outcomes in part because they increase exposure to stress and adversity 197. The inverse relationship between SES and cardiovascular risk is consistent across countries, in the Whitehall study and in the London cohort of the WHO multinational study of vascular disease in diabetes 198, and among women participants of the Alameda 24

39 county study in US 199. Mackenbach et al 200 found a similar relationship among 22 countries in all parts of Europe. In the Australian Mater-University of Queensland Study of Pregnancy, CHD risk factors at age 14 clustered greater among those from families with low income 201. One of the proxy for socioeconomic position is an individual s educational level 202. Lower educational attainment increased the odds of having a BMI greater than or equal to 25 kg/m 2 for both sexes (P=0.03 males and P<0.001 females) 36. Overweight and obesity were higher in rural females with lower educational achievement (P<0.05) and the socio-economically disadvantaged (P<0.0001) 203. In Australasia, hazard ratios for primary compared to tertiary education were 2.47 vs 1.47 for CVD mortality and 2.09 vs 1.23 for all CVD, independent of traditional CVD risk factors 204. b) Maternal Smoking in Pregnancy Tobacco smoking is an established major risk factor of CVD 205 and a leading cause of morbidity and mortality 206. The two facets of maternal prenatal smoking are causality related to nicotine exposure and adverse lifestyle behaviour of smoking per se. There is evidence to suggest that epigenetic patterns are altered by nicotine exposure 207. With regard to the former, maternal prenatal nicotine exposure prompts DNA methylation and likely modifies dietary fat intake preferences among offspring 208. The differences in infant DNA methylation associated with maternal prenatal smoking have been detected in childhood blood samples at ages 3 5 years 209. The effect of nicotine exposure on the in-utero development of hypothalamic function can also exert an impact on appetite control and energy expenditure throughout the life course 210,211 and promote obesity by enhancing dietary preference for fat through a neural mechanism involving the amygdala 212. There are epidemiological studies that suggest a causal relationship between prenatal adverse exposure, including maternal prenatal smoking and offspring obesity There is an argument against intrauterine mechanism(s) related to nicotine intrauterine exposure. The argument was invoked by a study that examined the relationship between maternal prenatal smoking and childhood BP 215 that found a similar association between offspring BP and maternal prenatal smoking or partner s smoking during the same period. The finding suggested that childhood BP was not influenced by intrauterine effects. The modest difference in systolic BP (unadjusted) between children of mothers that did and did not smoke during pregnancy (β=0.64 mm Hg; P=0.02) was no longer 25

40 significant after adjustment for breastfeeding, maternal education, and family social class. Thus, the differences in child BP observed in minimally adjusted models may not be attributable in totality to a biological influence of maternal smoking related in-utero environment 215. A review of epidemiological studies of maternal smoking and blood pressure presented inconsistent (both positive and null) results for children 216. A study of Swedish men conscripted for mandatory military service between 2001 and 2006, had young Swedish men born between 1983 and 1988 with information on both maternal smoking during pregnancy and blood pressure at military conscription. The cohort included 780 full brothers discordant for maternal smoking. The Swedish study results found a small but significant increase in both SBP and DBP for young men whose mothers had been daily smokers during pregnancy compared with sons of non-smoking mothers 217. Within-sibling analysis comparing 780 brothers discordant for maternal smoking exposure however found similar estimates, with no significant difference, after adjustment for age at conscription, height, BMI, parental cohabitation status at time of pregnancy, parity and maternal blood pressure disease during pregnancy. In summary, the study results reflected uncertainty of maternal smoking exposure and offspring BP association. This study had a major limitation of selection bias favouring men who had parents with higher education and mothers who less often smoked while pregnant, indicating a higher socioeconomic background. No associations were observed in a population-based birth cohort for maternal, paternal or parental smoking and BP levels of the offspring at ages 17 and The latter Jerusalem Perinatal Study is an example of study that examined the effect of maternal passive smoking during pregnancy and offspring BP at ages approximating young adulthood. Among 732 young adults from the Netherlands cohort that examined parental smoking and vascular damage in offspring (mean age 28.4) there was no statistical difference in young adult systolic or diastolic BP among offspring of mothers that smoked during pregnancy compared to maternal prenatal non-smokers 218. Adverse familial factors, such as low maternal education and not living with the baby s father, may predominate women who are exposed to passive smoking in their homes 217. Beyond the in-utero nicotine exposure perturbations, and maternal prenatal smoking collinearity with psychosocial variables, such as breastfeeding, maternal education, and family social class, and unhealthy and familial behavioural patterns may explain more 26

41 consequential adverse psychosocial influence of maternal prenatal smoking in pregnancy on BP. Smoking behaviours generally are linked more closely with low income, younger maternal age, and lower maternal education 219. c) Offspring current alcohol consumption There is evidence of a positive association between excessive alcohol use and incident hypertension in adults 220. In 491 Caucasian Australian males aged 20-45y, average weekly alcohol consumption correlated with their SBP (r=0.18) 221. Offspring from risky or dysfunctional families may consume excessive alcohol 136. Alcohol misuse is the highest risk factor for disability-adjusted life years (7 0% overall, 10 5% for males and 2 7% for females) for young people aged years 222. Offspring alcohol consumption is clearly an important confounder to be considered for adjustment in assessment of any relationship between maternal prenatal stress and offspring BP. Moreover, reducing alcohol consumption, at least in those who consumed 3 or more drinks per day, decreases SBP and DBP 223. The study examining the relationship of usual current alcohol intake with systolic and diastolic pressures among young adults (316 men and women, aged 18 to 26 years, from East Boston in United States) showed a J-shaped association of alcohol intake with blood pressure level in young adults, with the lowest levels in consumers of 1 to 3 drinks per day 224. Mendelian randomization has provided robust evidence on the nature of association between alcohol intake and blood pressure in adult males using a common polymorphism in aldehyde dehydrogenase 2 (ALDH2) as a surrogate for measuring alcohol consumption 225. ALDH2 encodes a major enzyme involved in alcohol metabolism. A fixed effect meta-analyses of the ALDH2 genotype with blood pressure (five studies, n = 7,658) and hypertension (three studies, n = 4,219) in males obtained an overall odds ratio of 2.42 (95% CI , p = ) for hypertension comparing wild-type homozygotes (*1*1) with null variant homozygotes (*2*2) and an odds ratio of 1.72 (95% CI , p = 0.006) comparing heterozygotes (surrogate for moderate drinkers) with *2*2 homozygotes 225. Individuals homozygous for the null variant (people who inherit the variant form of this gene from both parents) experience adverse symptoms when drinking alcohol and consequently drink considerably less alcohol than wild-type homozygotes or heterozygotes 225. There was no association between ALDH2 genotype and hypertension among the women in these studies (mainly from Japan) because they drank very little. 27

42 1.3 CVD risk factors- high BP and obesity- in context of Developmental Origins of Health and Disease (DOHAD) Background Increasingly, adult health disorders and cardiovascular disease (CVD) risk factors including abdominal obesity and high blood pressure (BP) are thought to be influenced by prenatal exposures. The growth and development of the foetus have been shown to be profoundly influenced by prenatal maternal cigarette-smoking 226 and alcohol consumption in pregnancy 227. These adverse prenatal environmental exposures may lead to the developmental programming of the foetus. The DOHAD theory suggests that the maternal environment has an important role in gestation 228, and factors including prenatal food inadequacy during wars or famine and maternal stresses comprise an adverse prenatal environment. The earliest ecological study to suggest that events in utero are important predictors of health outcomes later in life was undertaken in the UK in different parts of England and Wales 229. The differences in the death rate from CVD in those areas paralleled previous differences in their death rates among newborn babies, thereby linking the geographical difference in CVD death rates to impaired foetal growth 230. Three other common causes of adult death- chronic bronchitis, stomach cancer and chronic rheumatic heart disease- had a similar relationship with past infant mortality. These diseases are also linked to poor social conditions, strengthening the suspected epidemiological link between adverse environmental influences in utero and adult health disorders, including CVD. Other studies have supported the argument that CVD is linked with adverse influences in early life. For example, arteriosclerotic heart disease has been correlated with past infant mortality in the 20 counties of Norway 231. In 17 states of the US, mortality from CVD has been linked to infant mortality resulting from diarrhoeal disease, a finding consistent with that of the Norwegian finding 232. A study of civil servants in the UK showed that death rates were higher in those who were shorter in stature, and who may therefore have had an adverse environment, in terms of socioeconomic status (SES), in early life 233. Among long-term employees of the Bell System Company in the US, men whose parents had been in white collar occupations had a lower incidence of CVD than those from blue collar families

43 Undernutrition in utero could permanently change the body's structure, physiology and metabolism, and lead to CVD in adulthood. The in utero period is seen as a critical period when a system is plastic and sensitive to the environment, followed by loss of plasticity and a fixed functional capacity 229. A poor foetal environment provokes an adaptive response (the thrifty phenotype hypothesis) which optimizes the growth of key body organs to the detriment of others 235. The adaptive changes that seem to be beneficial for survival in the near-term could rapidly turn into maladaptation, modified by catch-up growth or diet induced obesity, and presenting an altered phenotype in later life 236. Such predictive adaptive response possibly accounts for extreme birth weights, either very low or very high birth weight, both scenarios predisposing the offspring to increased fat mass and incidence of metabolic syndrome in childhood and adulthood 237 Table 1.4 lists some of the major studies that heralded support for the DOHAD theory during the 1980s and onwards. The tabulated epidemiological studies focussed on CVD outcomes with relation to participant SES, birth weight (BW), change in BMI (catch-up growth) during childhood and adolescent period. It was considered more appropriate to present the data in chronology of epidemiological studies initiated by Barker et al towards the development of the DOHaD theory. In particular, early population based studies focused on association between birth weight or placental weights and long-term development of high blood pressure. Importantly, birth weight or placental weights were considered as surrogates of intrauterine growth of foetus in the hindsight of population cohort exposures to famines or wars). Examining studies in their chronological order makes it clearer how and why biological predictive adaptive response concept replaced Barker s thrifty phenotype hypothesis overtime in DOHaD theory (the theory suggesting an important role of maternal environment in gestation). From 1989 to 1993, Barker et al conducted four retrospective studies in the UK. Two out of the four studies measured an inverse effect of BW on BP 238,239. In 1989, Barker et al 238 found (from the national samples of y and 3259 adults) that both at 10y and 36y, systolic BP (SBP) inversely related to birth weight. At 10y, SBP increased in relation to current weight. In 1990, Barker et al 239 found (among 449 men and women born in ) an inverse relationship between birth weight and adult BP, independent of duration of gestation, cigarette smoking, alcohol consumption and social class currently or at birth. In contrast, BP showed a rise in relationship to a higher placental weight. The authors suggested that discordance between placental and foetal size may lead to circulatory adaptation in the foetus, altered arterial structure in the 29

44 child, and hypertension in the adult. The results of the two studies reiterated that geographical differences in mean BP and CVD mortality partly reflected past differences in the in-utero environment. The other two studies reported an inverse effect of BW on standardised mortality rate (SMR) among men 240,241. In 1989, Barker et al 240 demonstrated standardised mortality rates fell faster in men who weighed higher at birth, independent of social class adjustment. Thus, a large fall in CV mortality in the US, Canada, Australia, and New Zealand at that time (and during the past 20y) was attributed to improvement in child growth and health, reflected in the fall in infant mortality sixty and more years ago 242. Indeed modelling the decline in coronary heart disease deaths in England and Wales ( ) showed that primary prevention such as reduction in smoking, total, cholesterol, and a fall in mean population blood pressure resulted in fewer deaths 243. In 1993, Barker et al 241 reiterated the association between higher birth weight and a lower SMR. In the same year, Osmond et al 244 demonstrated that the inverse relationship between CVD and BW was similar in men and women and found that in men CVD was related to weight gain in infancy. Thus, the four study findings were supportive of opinions suggesting that the reduction in growth begins early in gestation, maternal nutrition affected programming, and that programming of the body's structure, physiology, metabolism by the environment during foetal life was the origin of CVD. In 1996, a review of 34 studies published between 1956 and 1996 showed that a fall in BP with increasing BW among adults 245. The effects of birth weight and early childhood weight gain were independent, so that adults who had been the lightest at birth but gained most weight in early childhood had the highest blood pressures. In 1997 Uiterwaal et al 246 showed the importance of low BW combined with high current body mass index (BMI) in the development of high BP. In this follow-up study of Dutch children aged 5 through 21 years, they found BW was inversely associated with SBP from childhood to young adulthood, and with DBP in young adulthood, but BW is not related to change of BP with increasing age. In 2000 Huxley et al 247 reviewed papers published between March 1996 and March 2000 that confirmed an inverse relationship between BW and BP was applicable to adolescents (a period in childhood when hormonal adjustments are peaking). However, the BW and BP relationships were attenuated compared to both the pre- and post-adolescence. 30

45 Some of the tabulated studies adjusted their findings for behavioural and lifestyle factors such as alcohol-use, smoking, and oral contraceptive (OC), gender and current height, current weight. A follow up study of women and men (from Hertfordshire, England) whose birth weight and weight at 1 year of age had been recorded, showed the relationship between cardiovascular disease and birth weight are similar in men and women 244. In neither women nor men were birth weight or weight at 1 year associated with death from lung cancer, which serves as an indicator of cigarette smoking 244. This way cigarette smoking did not affect the inverse relationship, linking birth weight to death from cardiovascular disease in adulthood. Uiterwaal et al 246 found a strong and consistent inverse association between birth weight and systolic blood pressure among the longitudinal cohort of 483 children from Netherlands. This inverse association persisted from adolescence into adulthood after adjustment for, particularly, current body height and weight besides the preliminary adjustments for sex, and use of alcohol, cigarettes, and oral contraceptives. The Nurses Health Study 248 comprised a longitudinal cohort of female nurses followed since 1976 who were born of singleton, term pregnancies and reported their birth weight in The study results showed for each kilogram of higher birth weight, age adjusted hazard ratios from prospective analysis were 0.77 (95% confidence interval 0.69 to 0.87) for coronary heart disease. Risk of coronary heart disease was especially high for women who crossed from a low centile of weight at birth to a high centile of body mass index in adulthood suggesting a higher BMI in adulthood is an especially strong risk factor for CHD among women who were small at birth. 31

46 Table 1.4. Major DOHAD epidemiological studies (late 1980s to date) Author (et al) Date Published Population Location and design Outcome Barker National samples of y-olds and 3259 adults Barker men born in Barker men and women born in Barker men born during Osmond women and men born during Law More than subjects age 0-71 Uiterwaal males and 231 females Eriksson men born during Huxley men and women aged 0-84 Cheung men and women born in 1967 Barker men born in Law men and women aged 22 Keller patients aged vs. 10 normotensives Rich- Edwards adult female nurses Barker men and women aged ~62 Britain; Retrospective cohort study Britain; Retrospective cohort study Britain; Retrospective cohort study Britain; Retrospective cohort study Britain; Retrospective cohort study Review Netherlands; Prospective cohort study Helsinki, Finland; Retrospective cohort study Review Hong Kong; Longitudinal study Helsinki, Finland; Retrospective cohort study Britain; Longitudinal study Germany; Case-control study America; Longitudinal study. Helsinki, Finland; Retrospective cohort 32 In boys, at 10y SBP by 0.38 mmhg (95%CI -0.04, 0.80) from lowest to highest BW. In girls 1.32 (1.03, 1.61) Standardised mortality rate fell from 111 in men who weighed 8.2kg at 1y to 42 in those who weighed 12.3 kg SBP & DBP rose by 15mmHg with 450g to >680 g rise in placental wt. BP fell by 11mmHg with 5.5lb to >7.5lb rise in BW SMR fell 119 in men with BW of 2495 g to 74 in men with BW of > 3856 g (p=0.02) Among women and men, SMR for CVD fell with BW. Women with BW 5.5lb had raised death rates, (SMR=120; 95%CI 85, 165) In adults, BP fell with BW. In neonates, there was a positive relationship between BP and BW Inverse association between BW & SBP, (-2.4 mmhg/kg; CI: -3.9, - 1.0). No relation between BW & change in SBP/DBP with age HR CHD death rate 14% (8-19) per unit (kg/m 2 ) in PI at birth. HR by 22% (10-36) per unit in BMI at 11 years of age BP fell with BW. The size of effect ~2 mmhg/kg. BP with head circumference. The size of effect was ~0.5 mmhg/cm Post-natal change in PI (β=-2.2) was inversely associated with SBP. Birth length associated with DBP (β=-2.6) HR in men with lowest annual income was 1.71(1.18, 2.48). HR was higher 2.58 ( ) with PI< 26kg/m 3 ) SBP by 1.3 mmhg (0.3, 2.3) for 1SD score in BW and, independently, by 1.6 mm Hg (0.6, 2.7) for 1SD score in early childhood wt. gain Patients with hypertension had fewer glomeruli/kidney than controls (median, 702,379 vs. 1,429,200) For each kg. of higher birth weight, age adjusted CHD HRs were 0.77 (95%CI 0.69 to 0.87) OR=1.6 (1.1, 2.3) for placental Wt. <550g vs.>750g and 2.2 (1.5, 3.3) in people whose fathers

47 study were labourers vs. upper middle class families Unnatural events have an uncanny ability to randomise the distribution of prenatal adverse exposure among populations including pregnant women in various stages of pregnancy. Randomisation of the exposures in turn ensures validity of the study findings. This approach has been utilised by the researchers to assess the effect of prenatal exposure to malnutrition at various critical periods of gestation in relation to BP and obesity. Survivors of the Dutch famine of had their BP measurements taken at age 50 years. During the time of famine, participant mothers had rationed caloric intake from protein, fat and carbohydrate, all proportionally reduced. The study investigators found that the high BP among the adult participants was linked to reduced foetal growth, but not to prenatal exposure to a balanced reduction of macronutrients in the maternal diet. In other words, there was an inverse association between foetal growth and blood pressure but no association of exposure to famine during gestation (compared with nonexposed participants that were born before or conceived after the famine). For instance, SBP differed by 1.5 mmhg (95%CI: -1.9, 4.8) among those exposed in late gestation, by -0.9 mmhg (95%CI: -4.2, 2.4) among those exposed in mid gestation, and by -3.6 mmhg (95% CI: -7.5 to 0.4) among those exposed in early gestation 255. Given that the exposure to famine among participants of this Dutch famine study was brief (5 months versus more than 2 years in Leningrad) 255, the effects of malnutrition during a short period of gestation on fetal growth may therefore have been buffered by nutritional reserves built up in the mother's body. This study reiterated the finding of inverse relationship between birth weight and BP. An increase of 1kg in birth weight was associated with 3.9 mmhg (95%CI: 1.2, 6.6) decrease in SBP, after adjustment for gestational age at birth, maternal (age, primiparity, weight at the end of pregnancy, weight gain in third trimester, SBP, manual labour) and adult characteristics (age at measurement, BMI, SES, current smoker, alcohol, antihypertensive medication). Similar to the Dutch famine study, the risk of hypertension development in later life was higher in severely affected famine areas of China, compared to those areas not exposed to famine 256. The affected adults had SBP and DBP difference of 2.2 mmhg (95%CI: 1.3, 3.0) and 0.9 mmhg (95%CI: 0.3, 1.5), respectively and higher odds of hypertension 33

48 OR=1.88 (95%CI: 1.00, 3.53) independent of age, sex, SES, lifestyle, dietary factors, and family history of hypertension. Ravelli et al 257 examined the effect of prenatal exposure to famine on the development of obesity in adult life in the historical cohort study of 300, year-old men exposed to the Dutch famine of Figure 1.4 shows various sub-cohorts according to prenatal timing of exposure and their obesity rates, compared to controls. In participants that were exposed to famine in the last trimester of pregnancy and the first three to five months of life (Cohort B1), the obesity rate was significantly lower than in the unaffected control population (0.82% vs. 1.32%; P<0.005). The cohort that was exposed to famine during the first two trimesters of pregnancy (Cohort D1) had a significantly higher rate of obesity than the control cohort (2.77% vs. 1.45%; P<0.005). These findings bear on the general hypothesis that conditions prevailing during the critical periods of prenatal growth may have permanent effects upon a given organ or tissue. The last trimester of gestation and the first months of life constitute a critical period for subsequent obesity. Dietary conditions prevailing during those critical periods are thought to determine the ultimate number of adipose cells laid down. Figure 1.4. Obesity prevalence rates among birth cohorts in famine and control. Adapted from: Ravelli et al (1976)

49 1.3.2 Prenatal life stress and DOHAD Prenatal stress exposure represents an adverse environmental condition that may contribute to variation in both birth phenotype and the physiology of the developing organism The historical studies that examined the relationship between nutritional inadequacies (due to wars, famines, natural disasters, or poverty) and BP or obesity did not delve into the possible intrauterine stress exposure interaction with the prenatal nutritional milieu. The long-term effects of intrauterine stress may target the effects of stress on foetal brain differently through maternal-placental foetal endocrine and immune/inflammatory mechanisms 261. Entringer et al 262 demonstrated the impact prenatal life stress exposure can have on offspring cardio-metabolic function. They conducted a case-control study among healthy young adults (age 25 ± 5.14 years) born to mothers with healthy pregnancies, one half born to mothers who had experienced a major stressful life event during the index pregnancy (n=36) and the other half was a socio demographically-matched population. Young adults exposed during intrauterine life to maternal psychosocial stress exhibited a significant alteration in glucose-insulin metabolism, endocrine function and immune function. These individuals had a higher BMI (24.6 kg/m 2 vs.22.5; P=0.04), higher leptin (768.2 vs. 391pg/mL), lower HDL-cholesterol (50.4 mg/dl vs. 59.8; P=0.04), and a higher insulin (P=0.046) levels. Leptin resistance, in addition to insulin resistance, is a critical endocrine defect in the pathogenesis of programming-induced obesity and metabolic disorder 262. Foetal exposure to excess glucocorticoids may play an important role in the foetal origins of adult disease 263. Glucocorticoids serve as potent regulators of foetal growth and development by altering the expression of many proteins at both the molecular and cellular level 264. Long-term alterations in the hypothalamic-pituitary-adrenal (HPA) axis are accomplished by hormonal signals, epigenetic modifications, and mitochondrial function 265. Experimental studies have shown that exposure to excess stress hormones during foetal life is associated with several physiological pathways that can be linked to future obesity via glucocorticoid programming 261,266,267. The exposure of the developing foetus to glucocorticoids, was proposed nearly two decades ago, following the observation that the foetus was protected from high maternal glucocorticoid levels by the actions of the placental barrier enzyme, 11β-hydroxysteroid dehydrogenase type 2 (11β-HSD2). The placenta is an effective barrier between the 35

50 maternal and foetal hormonal environments in humans and is rich in the above protective enzyme, which converts cortisol to inactive products such as cortisone 268. Glucocorticoids in the maternal blood circulation can, however, overcome the barrier by inactivating the 11β-HSD2 enzyme, allowing dexamethasone to cross the placenta and alter the foetal HPA axis Though the maternal HPA axis becomes gradually less responsive to stress as pregnancy progresses, an increase in cytokines, in response to stress, can facilitate an increase in transfer of maternal cortisol across the placenta to the foetus partly as a result of reduced placental 11β-HSD It remains to be elucidated whether such mechanisms operate in humans 273, although the differences in cortisol in pre-adolescent children (in a controlled environment), based on their prenatal anxiety exposure, is indicative of lasting effects on offspring HPA axis functioning 274. Children whose mothers had higher levels of morning cortisol during pregnancy and higher levels of stress about bearing a disabled child had higher levels of cortisol on school days compared to weekend days 275. A positive correlation between hair cortisol levels in mothers and their children, suggests a heritable trait 276. Cortisol assessment from maternal hair is akin to a geological survey, indicating the period of excess exposure to cortisol during their pregnancy. Maternal life stress in the prenatal period can invoke developmental programming of human health for offspring obesity and hypertension 261,266,267,272. Therefore, prenatal period care is important in the scheme of adult-onset disease prevention in the context of obesity and raised BP

51 1.3.3 Review of the evidence on the relationship between maternal distress in pregnancy and offspring BP Table 1.5 lists three major contemporary pregnancy-cohort studies that have examined the relationship of maternal stresses during pregnancy and offspring BP development. A brief description of each study, the main findings and limitations are presented, followed by a critique of evidence on the relationships between maternal distress during pregnancy and offspring BP in childhood or adolescence. Table 1.5. Three contemporary pregnancy cohort studies examining maternal distress during pregnancy and offspring BP Author (et al) Date Publish ed Population Analysed Location and design Outcome van Dijk 278 (ABCD study) offspring aged 5-7 Amsterdam, Netherlands; Pregnancy cohort study Cumulative stress scores of 3-4, significantly associated with both SBP (1.5; 95%CI: 0.0, 3.0) and DBP (1.5; 95%CI: 0.1, 2.9) Taal 279 (Gen-R study) van Dijk 280 (ALSPAC study) offspring aged offspring aged Rotterdam, Netherlands; Pregnancy cohort study Britain; Pregnancy cohort study *Confounder adjustments are listed in Table 1.6 Maternal psychological distress in pregnancy associated with SBP (0.86; 95%CI: 0.11,1.62), DBP (0.72; 95%CI: 0.08, 1.36), adjusted for sex and age Maternal depressive symptoms and DBP are inversely related after adjustment for confounders*(-0.084; 95%CI: , ) a) Amsterdam Born Children and Their Development (ABCD) Study The large prospective population-based ABCD study established in 2003 examined the association between maternal lifestyle, medical, psychosocial and environmental conditions during pregnancy and children s health at birth and later life 278, mother-child pairs (37.7% follow-up of live born singletons) were included in the study. Mothers filled out the pregnancy questionnaire around mean gestational week 16 and the children s BP was measured at 5-7y. In the ABCD cohort analysis, multiple inventories were administered to ascertain stress and anxiety during pregnancy. The validated Dutch version of the 20-item Center for Epidemiological Studies Depression Scale (CES-D) questionnaire was used for determining depression during pregnancy. The results of the ABCD Pregnancy study showed that prenatal presence of multiple psychosocial stressors was associated with higher SBP (Coefficient = 0.04; 95%CI: 37

52 0.00, 0.07) and DBP (coefficient=0.06; 95%CI: 0.02, 0.09) in the 5-7y children. These results were minimally adjusted for sex, height and child age. Once adjusted for more confounders, the associations were no longer significant. The adjusted confounders were maternal age, ethnicity, pre-pregnancy weight and height (BMI, kg/m 2 ), educational level, primiparity, smoking, alcohol consumption, gestational age, birth weight, child BMI, and pre-existing conditions in pregnancy (including hypertension, diabetes, hypothyroidism, hyperthyroidism and epilepsy). The cumulative prenatal stress scores of 3-4 were positively associated with both SBP (coefficient=1.5; 95%CI: 0.0, 3.0) and DBP (coefficient=1.5; 95%CI: 0.1, 2.9), but for the hypertension outcome the association was not significant after adjusting for above confounders. Overall the ABCD study showed after adjustment for confounders, the single stress scales (depressive symptoms, state-anxiety, pregnancy-related anxiety, parenting daily hassles and job strain) were not associated with systolic and diastolic BP, or hypertension. However, cumulative stress (denoted by the presence of 3-4 psychosocial stressors prenatally) was significantly associated with a higher systolic and diastolic BP in childhood compared to no stressors. b) Generation-R Study Another population-based prospective cohort study was the Generation R Study undertaken in the Netherlands. This study was designed to identify early environmental and genetic causes and causal pathways leading to normal and abnormal growth, development and health during foetal life, childhood and adulthood. In total, 9778 mothers with a delivery date from April 2002 until January 2006 were enrolled 279,282. A nested cohort of the Generation R study comprised children (n=4831) aged 6. The association between offspring BP and maternal and paternal psychological distress was assessed in the second trimester of pregnancy, using multiple inventories. A score was provided for each symptom scale by summing the item scores of each scale and dividing the results by the number of endorsed symptoms. The resulting scores were from 0 to 4 with higher scores representing an increased occurrence of symptoms of depression and anxiety. Overall maternal psychological distress during pregnancy was associated with childhood SBP (coefficient=0.86; 95%CI: 0.11,1.62) and DBP (coefficient=0.72; 95%CI: 0.08, 1.36), after adjustment for child sex and age at measurement. However, after additional adjustment for maternal age, parity, educational level, smoking habits during pregnancy, ethnicity, pre-pregnancy BMI, age, 38

53 BP at intake, gestational age at birth, birth weight, breastfeeding and BMI at age 6, the association was no longer significant for SBP or DBP. One of the limitations of this study is that the information about maternal psychological distress was missing for 24.5% of all mothers, and for 32% of all fathers, of whom the partners had available data on maternal psychological distress. The authors identified that bias could arise mainly from loss to follow-up rather than non-response at baseline, as was the case in their study. The mothers from children who did not visit the research centre, had higher psychological distress scores, smoked more frequently and were less well educated than the total sample. Thus, a selective loss to follow-up might have led to biased effect estimates and a loss of statistical power. These findings did not support the hypothesis that maternal psychological distress during pregnancy is associated with childhood BP. Taking the varied results of the two Dutch studies into account it is safe to pronounce that a multivariable adjusted relationship in the ABCD study, between the maternal prenatal stress and childhood BP, could not be validated in a similar aged Gen-R cohort. c) The Avon Longitudinal Study of Pregnancy and Childhood (ALSPAC) Study The ALSPAC 280 is a birth cohort study of all pregnancies amongst women resident in three English health districts with expected delivery dates between 1 April 1991 and 31 December The study recruited pregnancies leading to surviving children. The study aims to follow each child through to adulthood. The eligible cohort for the analyses at the 10-11y follow-up included 4318 mother offspring pairs. Complete data on maternal depressive and anxiety symptoms at 18 and 32-week gestation, offspring BP (obtained at the 10-11y follow-up clinic), and other demographic data were available for all mother-offspring pairs. The study used multiple inventories to ascertain prenatal maternal and paternal anxiety, and postnatal depression. To compare prenatal and postnatal maternal stress they categorized all Edinburgh Postnatal Depression Scale and Crown Crisp experiential index scores into high levels of anxiety/depressive symptoms (yes versus no) by dividing the 80th percentile into a high group and a normal group. Gender difference in the association was tested by an interaction between sex and exposure (depressive/anxiety symptoms). No strong statistical evidence of interaction 39

54 by sex was found. Therefore, the results were presented with both females and males combined. Maternal depressive symptoms at 18-week gestation were inversely associated with lower offspring DBP but no association with SBP (Table 1.6). At both gestational ages of 18 and 32 weeks, there was no strong evidence that maternal depressive symptoms or anxiety were associated with adverse cardiovascular function in any of the multivariable models. Table 1.6. Relationship between prenatal depressive symptoms and BP. Maternal depressive symptoms at gestational week 18 Systolic BP Diastolic BP β (95% CI) β (95% CI) Model 1: Adjusted for offspring sex, height, age (-0.096, 0.021) (-0.126, ) Model 2: As model 1 plus adjustment for maternal age, ethnicity, pre-pregnancy BMI, nulliparity, social class, smoking, alcohol consumption. Model 3: As model 2 plus adjustment for gestational age and birthweight (-0.111, 0.008) (-0.136, -0.03) (-0.110, 0.009) (-0.135, ) Model 4: model 3 plus adjustment for BMI child (-0.112, 0.006) (-0.136, ) Adapted from: van Dijk et al (2013) 280 Though the inverse association of maternal stress with offspring DBP was different to the paternal null association, offspring of mothers with postnatal depressive symptoms or anxiety had a small but lower SBP and DBP, not explained by gestational duration, birthweight, or offspring age. Overall, they found no evidence to support the hypothesis that maternal stress during pregnancy adversely affects offspring vascular function at age via intrauterine mechanisms 280. The direction of a positive relationship between maternal stressors in pregnancy and offspring BP (aged around 6) was similar in the two Dutch studies. The relationship in the Generation R study was, however, not independent of multivariable confounding. The ALSPAC study children (aged 10-11) showed an inverse relationship between prenatal stress exposure and DBP. While the presence of multiple psychosocial stressors prenatally, in the ABCD study, was associated with 1.5 mmhg higher offspring SBP and DBP, their presence did not show a significant increase in the risk for hypertension. This may be related to the non-uniformity in the distribution of maternal stressors for the two levels of dichotomous hypertension status or the sample size was small in the hypertension group, given a small prevalence of hypertension in the population at age 6. Additionally, the reference values for defining hypertension obtained by the auscultatory method may not have been as precise compared to the oscillometric 40

55 assessment of BP in reducing the potential inter-observer differences. The major difference between the findings of the ABCD and the ALSPAC study was the direction of the association between the prenatal stress exposure and offspring BP. Whereas the association between prenatal psychological distress and BP, after adjustment for age and sex was positive in the ABCD study, the relationship between prenatal psychological distress and DBP was inverse in the ALSPAC study. Given that the ALSPAC study did not measure the cumulative effect of multiple psychosocial stressors, unlike in the ABCD study, a true comparison cannot be drawn between the two studies. Also, the age-groups in the two studies were different, making the direct comparison of their results difficult. Minimally adjusted models among the two studies from the Netherlands (the ABCD study and Generation-R study) showed a positive relationship between maternal distress and BP. However, following multivariable adjustments the associations were statistically non-significant. It is evident from the results of the three pregnancy cohort studies that the necessity of multivariable adjustments, over and above the minimal adjustments for age and gender, is paramount to draw any valid inference about the association. Multivariable adjustments ranging across psychosocial, behavioural lifestyle and phenotypic factors are useful in drawing unbiased inferences from the statistical models. The fact that participants in the ALSPAC study were nearing the age of adolescence means the outcome of this study such as an inverse relationship between prenatal depressive symptoms and BP (DBP) required more attention, contrasting the positive direction of the age- and sex- adjusted association between prenatal psychological distress and offspring SBP and DBP in the ABCD study at age 6. 41

56 1.4 Gaps in knowledge There are many studies demonstrating relationships between adult hypertension and low birth weight or birth size, but paucity of prospective cohort studies showing an association between maternal stressors during pregnancy and offspring BP or BMI at age 20. Moreover, no study has examined longitudinal relationships between prenatal stress and BP and BMI, from childhood through adolescence to young adulthood. Recently, three contemporary pregnancy cohort studies have conducted cross-sectional examinations between maternal distress during pregnancy and BP at age 5-7, 6, and 10-12, respectively. However, the three studies did not examine for the effect modification of childhood adiposity on the relationship between maternal distress and offspring BP. The rising trends in children adiposity envisages an assessment of the difference in the magnitude and direction of the prenatal stress exposure and BP relationship according to various adiposity levels, that represents a gap in the three pregnancy cohort studies. Confounding can arise because of interrelation between the socio-demographic, phenotypic, psychosocial, or lifestyle factors. Confounders are extraneous factors that wholly or partially account for the observed effect of the risk factor on health outcomes such as hypertension or obesity. While the issue of confounding had been considered a priori in most of the reviewed studies that examined the relationship between depression symptoms and BMI, ascertainment of underlying interactions was not carried out. Accounting for effect modification of phenotypic bio-behavioural, psychosocial, or lifestyle risk factors on the relationships between BMI and depression symptoms may yield more realistic estimates. None of the studies that studied the relationships between BMI and depression symptoms adequately took account of underlying influence of psychosocial or familial factors on the association between the offspring depression and adiposity. Maternal prenatal smoking and low family income are important maternal and socio-economic factors that identify individuals vulnerable to the coexistence of obesity and depression in emerging adulthood. The studies in this thesis examined whether a positive association between depression symptom score and BMI is related to maternal adverse lifestyle during pregnancy or an adverse circumstance of low family income at pregnancy. Residual confounding was also a common insufficiency in the epidemiological studies that examined associations between offspring depression and BP. For instance, an 42

57 absence of adjustment for contraceptive use especially among girls is notable in previous studies. Oral contraceptive use is an important lifestyle marker that has been positively associated with BP. Oral contraceptive use could partially account for the observed effect of depression on BP in females. Previous prospective studies examining the relationships between depression/anxiety and BP in young adults have reported contrasting results. The usual shortcomings identified for the inconsistent results are either insufficient follow-up time, inadequate sample size, or lack of well validated standardized measures. Differences between the results of studies may have been in part due to differences in the populations selected, age related factors, anti-depressant or anti-hypertensive treatment, and varying means of assessing depression or anxiety. Other important sources of heterogeneity include different outcome measures such as systolic BP, diastolic BP, BP changes, or hypertension. None of the studies examined moderating influences of adiposity on the relation between depression or depressive symptoms and BP, leaving uncertainty about the study results for the direction and magnitude of the associations. The final study will assess the interacting influence of offspring adiposity on the relationship between depression and BP at age 20. The magnitude and direction of the relationship between depression/anxiety symptom scores and BP will be ascertained, accounting for major life style confounders including female oral contraceptive use at age

58 1.5 Scope and aims of the thesis In view of the importance of young adulthood as a crucial time for early recognition and prevention of future CVD, this thesis aims to assess the influence of antenatal and postnatal psychosocial/familial determinants that underlie the relationships between young adulthood BP and BMI and prenatal life stress exposure and negative emotional states of depression/anxiety. Significant relationships, accounting for important biological and socio-behavioural factors including their interactions, could provide clues for potential predisposition to long-term CVD risk. Young adulthood is an important period in which both protective and adverse health behaviours are adopted and engrained, thereby, influencing long-term CVD risk. Despite decades of research, the optimal preventive measures to control high BP have not been established. Ageing is an important factor that has a particularly strong association with comorbidity/multi-morbidity Among 9156 patients from an Australian population who attended general practices in July November 2005, prevalence of multi-morbidity increased steadily with age, from 2.6% among the sampled people younger than 25 years to almost half of those aged years, three-quarters of those aged years, and four out of five of those aged 75 years or older 283. Young adulthood, unlike middle or late adulthood, is accompanied with minimal multi-morbidity/comorbidity. Emerging adulthood is a critical time for establishment of adult behaviours 286. The three-contemporary pregnancy cohort life course studies, that assessed the influence of prenatal maternal stress on adiposity and BP among children and adolescents, are yet to examine the association at emerging adulthood. The first study in this thesis examined the effect of prenatal life stress score on offspring BP and BMI at age 20, adjusting for a range of demographic and psychosocial or bio-behavioural lifestyle factors, including interactions. An adverse psychosocial determinant such as maternal prenatal smoking triggers weight gain and obesity in children 99,129 as does the adverse family determinant of low socioeconomic status 116. The studies that documented positive relationships between offspring obesity and depression or anxiety symptoms 63,98,287 did not account for potential interactions with adverse psychosocial and family determinants. This is highly relevant given an increase in the co-existence of depression symptoms and adiposity in young adults 63,97,98. Maternal prenatal smoking or a low family income at pregnancy 44

59 could have an influence on the linear relationship between depression symptoms and adiposity. The second study in thesis examined the underlying influence of psychosocial determinants such as maternal prenatal smoking or family income on the relationship between offspring depression symptoms and young adult BMI. Adults with depression or anxiety symptoms are more likely to be obese than adults without them and obesity per se is positively associated with BP 184,288. Obese adults are at a 1.2 to 1.5 times increased risk of CVD 37. Considering a high prevalence of coexisting depression and obesity in young adults, offspring BP outcome may vary at various levels of adiposity or depression. The prospective studies that examined the relationships between depression and BP in adults gave conflicting results and fell short of adjusting for an underlying influence of depression and obesity interaction on BP outcome. The third study examined the relationships between resting BP and depression/anxiety symptoms or self-reported diagnosed depression. 45

60 Chapter 2 The Western Australian Pregnancy Cohort (Raine) Study The Western Australian Pregnancy Cohort (Raine) Study is the source for research in this thesis. Prospective pregnancy cohort studies that follow the same participants from in utero through life are rare. The Western Australian Pregnancy Cohort (Raine) Study, began in Perth in May 1989 and has followed 2900 pregnancies from 2804 women and their 2868 live births, representing 96% of the initial recruitment sample since then. The original aim of the Raine study was to determine the effects of frequent ultrasound during pregnancy. The pregnant women were randomly assigned to either routine obstetric ultrasound (control group) or multiple scans (intensive group), mothers recruited from surrounding private clinics and the public antenatal clinic at King Edward Memorial Hospital, a tertiary perinatal centre in Western Australia 289. Women were invited to participate if they were between 16 and 20 weeks pregnant (average 18), could understand and speak English, expected to deliver their baby at King Edward Memorial Hospital, and intended to remain in Western Australia. Of those eligible to participate, 90% agreed to take part in the study. Extensive data were collected during pregnancy. The children were assessed at birth, and were subsequently followed up as a cohort at ages 1, 2, 3, 5, 8, 10, 14, 17, and 20 years using questionnaires and physical assessments. Written parental and adolescent/young adult consent (14, 17, and 20y) was provided at each follow-up. Questionnaire data, physical measurements and biological samples (ages 14, 17, 20y) determined growth, cardiovascular, respiratory, immunological, musculoskeletal, nutritional, psychiatric, neurocognitive and ophthalmic health 290. Anthropometric and clinical measurements collected at all time points enabled the use of longitudinal methods to examine offspring BP and BMI over time with adjustment for a range of psychosocial, biological, socioeconomic, familial, and life style determinants. Ethics approvals for each of the Raine Study cohorts were sought from the Human Research Ethics Committees at King Edward Memorial Hospital, Princess Margaret Hospital for Children and The University of Western Australia. The details can be found at The publications under the Raine Study can be accessed at The initial 46

61 Raine sample overrepresented socially disadvantaged families and selective attrition of the sample over time had led to a closer representation of those in the sample to the Western Australian population 291. The illustration depicting attrition of the sample over time is placed below. Figure 2.1. The Western Australian Pregnancy Cohort (Raine) Study participants from birth to the 20-year review. Adapted from: Straker et al (2017) 292 NB: Deferred means the participant was in the cohort but declined to participate in the current review and withdrawn means there was no further contact with the participant. 47

62 Comparisons between the Raine Study and the Western Australian populations are tabulated below. The Raine cohort young adults and Western Australian population young adults were not significantly different (significance determined by a marked proportional difference >10%) for employment and working hours a week 292 (Table 2.1). Table 2.1. Comparison of Raine Study cohort participants at age 20 with contemporaneous Western Australian (WA) Census population (2011 census data) Raine (%) WA (%) Family structure Not married De facto married Any children Education completed Secondary > year Post-secondary / = tertiary study Labour force/occupation Professional/ managerial a Clerical/retail a Technical/trade/ labour a Unemployed/Not in labor force Work hours per week < Income levels Low Medium High Adapted from: Straker et al (2017) 292 a Out of total number employed in labour force To assess any attrition bias, the characteristics at infancy of participants and nonparticipants were compared at 20-year follow-up (Table 2.2). The proportions of participants and non-participants across a number of infant characteristics remained constant at 20-year follow-up, with the exception of reduction in participation of infants of Aboriginal and Torres Strait Islander ethnicity

63 Table 2.2. Comparison of participants and non-participants at 20-year follow-up by infant characteristics at birth Gestational age: missing n = 11 (0.4%) Participant (No) Participant (Yes) n (1406) % (49) n (1462) % (51) Term (" 37 weeks) Premature Birthweight a < 100% % % Small for GA b Large for GA c High-risk birth d Ethnicity Caucasian * 85.5 Aboriginal or Torres Strait Islander Other Adapted from: Straker et al (2017) 292 a %of mean birthweight for gestational age based on Western Australian norms b Small for gestational age (GA): < 90% expected birthweight (based on Australian birthweight norms) c Large for GA: > 110% expected birthweight (based on Australian birthweight norms) d Emergency caesarean section *P < Raine Study data suggests that CVD risk factors start manifesting early in childhood. Burke et al 293 have shown that overweight/obesity rates increase from 3 years of age, and at 8 years, the prevalence in boys and girls was 15% and 20% respectively. Burke et al 293 found at 8 years children with overweight including obesity had systolic and diastolic BP higher by 6 mmhg and 2 mmhg, respectively. The authors also found HDL-cholesterol was lower by 8% (P=0.002), and triglycerides were higher by 27% (P<0.001). Burke 294 presented that clustering of risk factors (adverse dietary choices, smoking, sedentary behaviour and alcohol consumption) in obese children is common and adverse lifestyle choices tend to aggregate in families (including maternal smoking) and recognition of these behaviours offers opportunities to prevent obesity in children and improve risk in other family members. The average SBP of children aged 6, whose mothers smoked during pregnancy was 1.2 (95%CI: 0.0, 2.3) mmhg higher than that of 49

64 children whose mothers did not smoke during pregnancy and maternal smoking of > 20 cigarettes per day associated with an increase in SBP of 3.4 ( 0.5, 7.4) mmhg with a corresponding increase in DBP of 2.2 ( 1.6, 6.0) mmhg 295. Chivers et al 296 showed that overweight/obese adolescents at age 14 had a different BMI trajectory pattern since birth, relative to those with normal weight. The BMI rate of increase was higher for girls compared with boys. In another study, they found maternal education, parental birth weight, and parental BMI were the strongest influences on their child s BMI from birth to adolescence 130. Huang et al 297 delineated a quarter of the Raine study participants at age 8 into highrisk cluster with a higher BMI, a higher BP, more adverse lipid profile, and a trend to higher serum glucose resembling adult metabolic syndrome. The risk factors common to this high-risk cluster were increased likelihood of greatest weight gain from 1 to 8 years (OR=1.4; 95%CI: ) and maternal smoking in pregnancy (OR=1.82; 95%CI: ). In a subsequent study at age 14, the authors found relative to the low-risk cluster the high-risk adolescents had a significantly higher C-reactive protein (P<0.001), uric acid (P<0.001), alanine aminotransferase (P<0.001), and gamma-glutamyl transferase (GGT) (P<0.001) 298. Huang et al 299 also identified seven distinct adiposity trajectories in the adolescents at age 14. They found insulin resistance was greatest in adolescents with increasing adiposity trajectories. Huang et al 300, employed a two-step cluster analysis to identify 17-yr-olds at high metabolic risk. The high-risk metabolic cluster included 16% of males and 19% of females. Compared to the low-risk group, the high-risk cluster participants had greater waist circumference, triglycerides, insulin, and SBP and lower high-density lipoprotein-cholesterol (all P<0.0001). These data show sexual dimorphism in effects of early life body mass index and fat distribution upon cardiometabolic risk factors 300. The two metabolic clusters derived statistically involved use of arbitrary cut-offs, placing the data in compartments, thereby, underutilising a full range of population values. The type of food may have a role in the underlying inflammatory processes. A literature-derived, population-based dietary inflammatory index has recently been developed to compare diverse populations on the inflammatory potential of their diets 301. The dietary inflammatory index seems to be a useful tool to appraise the inflammatory capacity of diet and to better understand the relationships between diet, inflammation, and cardio-metabolic diseases 302. Ambrosini et al 303 found in the Raine 50

65 Study adolescents (at age 14) that higher scores for the Western pattern were associated with greater odds of being in the high-risk metabolic cluster group, the features of high-risk group akin to adult metabolic syndrome. The Western diet is high in red and processed meat, take-away food, fried and refined foods, confectionary, and full fat dairy products and Healthy diet is rich in legume and fish, fruits, vegetables and whole grain. Studies on the Raine study adolescents at age 17 examined the relationships between behavioural lifestyle factors such as smoking and HDL-cholesterol or inflammatory markers such as high-sensitivity C-reactive protein. Le-Ha et al 304 found among 1050 adolescent participants a positive relationship between active smoking and hs-crp, modified by girls oral contraception (OC) use. A more robust effect of smoking on hs- CRP was observed in girls not using OC, whereas smoking was not significantly related to hs-crp among OC-using girls or boys. A significant positive association was also found between HPA measures (cortisol and corticosteroid binding globulin) and hs- CRP and blood pressure 305. Given that oral contraceptive use augments angiotensin II (AngII) levels, skin AngII type 1 receptor (AT1R) mrna, and hemodynamic response to AngII 306,307, the nonsignificant finding of smoking on hs-crp among OC-using girls may seem unusual. The hemodynamic effects of OC-mediated renin-angiotensin system activation among many OC-users can, however, be modulated by the mediation of estrogen component (of the combined oral contraceptives) on the functional feedback balance that exists between both angiotensin (Ang) II and nitric oxide under normal conditions 308,309 In the Raine study, the proportion of obese girls at age 17 was higher in the non-oc group than those using OC. It could be that the effect of smoking on hs-crp in non-oc users is driven by underlying obesity. A higher adiposity positively contributes to an increased reninangiotensin aldosterone system activity 308. Le-Ha et al 310 showed that being exposed to passive smoke since birth was associated with reduced HDL-C levels at age 17 in girls but not boys. Among non-smoking adolescent girls (n=374), those exposed to passive smoking and no maternal prenatal smoking had an inverse relationship with HDL-C (-0.09; 95%CI: -0.15, -0.02). The findings were adjusted for a range of confounders. Once again, it could be that the effects of obesity and OC non-usage are indistinguishable, given a higher proportion of obese girls at age 17 were non-oc users. 51

66 Le-Ha et al 311 reported that behavioural (OC use in girls and alcohol consumption in boys) and phenotypic factors (adiposity) influenced BP levels among adolescents. At age 17, boys had 8.97mmHg higher SBP, as compared with girls. Girls using OCs had 3.27 and 1.74 mmhg higher SBP and DBP, respectively, compared with non-oc users. There was a positive relationship between BMI and SBP, but the gradient of the relationship was steeper in boys compared to non-oc using girls. In boys, SBP was 5.7mmHg greater in alcohol consumers who were in the upper quartile of BMI and the urinary sodium-potassium ratio quartile compared with teetotallers in the lowest quartile. In girls, SBP was 5.5mmHg higher in those taking OC, in the highest BMI and urinary sodium-potassium ratio quartile as compared to those not taking the OC pill and in the lowest quartile. The findings of this study reiterated the gender-related differences in the effects of adiposity on BP at adolescent age 17. Physical activity was associated with lower SBP at age 17. The three studies by Le-Ha et al indicate that confounding adjustments are important for unbiased estimations. However, caution needs to be exerted while testing multiple level interactions as that may increase the chance of type- I error. The association between psychosocial traits and CVD risk factors was also examined in the Raine study participants. Louise et al 170 found boys (aged 14y) with higher anxious/depressed scores had lower systolic BP trajectories. The predicted mean difference between the BMI of girls with an anxious-depressed score in the top 5% and those who had no symptoms were respectively 0.6 kg/m 2. Boys with more anxiousdepressed symptoms had higher triglyceride measurements (p=0.008). The predicted mean difference between SBP of boys with a depressive symptom score in the top 5% and those who had no symptoms was 3.0 mm Hg. Louise et al 312 showed a variation within or close to the Monoamine oxidase A gene may explain in part the association between lower depressive symptom scores and higher SBP in Caucasian boys within the Raine cohort. 52

67 Chapter 3 Contrasting effects of prenatal life stress on BP and BMI in young adults 3.1 Preamble Prenatal exposure to stress events possibly render alterations in the biological stress systems of the foetus that can have a long-term impact on the cardio-metabolic system. This study examined the effect of prenatal life stress score on BP and BMI at age 20. It highlights the potential impact of prenatal life stress exposure on BP, adjusting for important anthropometric and behavioural factors that influence systolic and diastolic BP levels in this age group. By demonstrating a variety of behavioural and phenotypic factors and their differential impact on BP levels, it also addresses a theme central to this thesis, that prenatal life stresses may invoke developmental programming of offspring BP and adiposity that can be detrimental for future risk of CVD. Published in the Journal of Hypertension. April 2015; Vol 33(4):

68 3.2 Abstract Various environmental stressors in pregnancy have been reported to affect high blood pressure in adult offspring. However, few studies have examined the effect of prenatal maternal psychological stress on offspring blood pressure and body mass index in early adulthood. In 957 Raine cohort participants regression analyses were used to examine the association between the number of maternal life stress events experienced during pregnancy and offspring blood pressure and body mass index, at age 20. Prenatal life stress associated positively with offspring body mass index but inversely with systolic blood pressure. After adjustment for confounders, each additional prenatal life stress event reduced offspring systolic blood pressure by 0.66 mmhg (P=0.013) in those with an average body mass index and lowered the odds of systolic (pre)hypertension by 17% (OR=0.83; P=0.008). The inverse relationship between prenatal life stress and adult SBP was stronger in offspring with higher body mass index. On the other hand, each unit increase in prenatal life stress score predicted a body mass index increase of 0.37 kg/m 2 (P=0.022). Longitudinal analysis showed similar effects of prenatal life stress for offspring body mass index from age 8 and SBP from age 14. This study has shown that maternal stress in pregnancy significantly associated with body mass index from early childhood, but contrary to our hypothesis predicted lower resting systolic blood pressure and lower odds of systolic (pre)hypertension in young adult offspring. The effect of prenatal life stress on BP was accentuated by a higher body mass index. Foetal programming events because of prenatal stress may underpin some of these relationships. Key words: Early Life Stress, Raine Study Pregnancy Cohort, Psychosocial, Blood pressure, Body mass index 54

69 3.3 Introduction Several types of human maternal stress in pregnancy have been associated with offspring obesity and associated metabolic disorder [1, 2]. Such studies include maternal starvation, as in the Dutch famine in the Second World War [3, 4], the Chinese famine [5], maternal exposure to the events of the Holocaust [6, 7], environmental disasters such as the Canadian Ice blackout [8], and psychosocial stress in the form of maternal bereavement [9, 10]. These human observational studies have been complemented by an extensive literature on a variety of experimental pregnancy stressors in animals leading to offspring obesity and associated metabolic disorders [11-13]. The association between extremes of birthweight and adult blood pressure (BP) has been well documented following original observations of Barker and others [14-18]. However, the few studies of the relation between commonly experienced forms of human psychosocial stress in pregnancy and higher BP in offspring have been restricted to children and with inconsistent findings [19-22]. Given the associations between severe maternal stressors and offspring obesity it might be expected that there would be similar associations between maternal psychosocial stress and offspring BP, mediated either via body mass index (BMI) or independent mechanisms such as foetal programming of the hypothalamic pituitary axis [23-27]. We have, therefore, sought to test the hypotheses that prenatal life stress will be associated with higher BP and higher BMI in a young adult population using data from a well characterized pregnancy cohort first studied when the mothers were 18 weeks pregnant. The Western Australian Pregnancy Cohort (Raine) study is one of the longest running prospective studies on child health and development [28]. To the best of our knowledge, no other longitudinal study has examined the association between prenatal stress events exposure on BP and BMI in young adults. 55

70 3.4 Material and methods This study was conducted using data from the Raine Study, a prospective longitudinal follow-up study of 2868 live births followed from All pregnant women attending King Edward Memorial Hospital (KEMH) were enrolled at 18 weeks of gestation between August 1989 to April 1992, with details published elsewhere [28]. The study is based on the 20-year follow-up of the offspring, conducted from January 2010 to April The Raine Study has ethics approval from the Human Research Ethics Committees at KEMH, Princess Margaret Hospital for Children or The University of Western Australia. Informed written consent was obtained from the adult participants. This analysis includes the data collected during pregnancy and when their offspring were young adults. Prenatal life stress events Antenatal women were asked in the 18-week gestation questionnaire whether they had experienced any of the ten life stress events since becoming pregnant and, in the 34- week gestation questionnaire, whether any of the events had been experienced within the last 4 months [29]. The response was yes or no to each of the life stress events relating to: 1) Pregnancy problems, 2) Death of a close relative, 3) Death of a close friend, 4) Separation or divorce, 5) Marital problems, 6) Problems with their children, 7) Own job loss (not voluntary), 8) Partner s job loss (not voluntary), 9) Money problems and 10) Residential move [29]. As a number of life stress events were related to commonly themed experiences, the ten life stress events were collapsed into six categories. The two life stress events related to death of a close relative or a close friend were collapsed into one and the two life stress events relating to separation or divorce and marital problems were taken together. Similarly, the mother s or partner s job loss and money problems were combined into a single life stress event reflecting money insufficiency. As a result, the maximum life stress events score for any participant was six. It was noted that although some life stress events were recorded at both 18 and 34 weeks gestation, the nature of the life stress events suggested they were likely to have occurred only once and the stress perceived was sustained. To remove potential double counting, the number of new life stress events at 34 weeks (previously unreported) was added to the sum of life stress 56

71 events at 18 weeks to obtain a total score of life stress events experienced from conception to 34 weeks pregnancy. Offspring blood pressure and anthropometric measurements BP and anthropometry were measured at age 1, 3, 5, 8, 10, 14, 17 and 20 years. Resting supine BP and heart rate were recorded using an oscillometric sphygmomanometer (Dinamap ProCare 100; Soma Technology, Bloomfield, Connecticut) with an appropriate cuff size. After resting quietly for 5 minutes, BP recordings were taken sequentially every 2 minutes. At age 14, 17 and 20 years the last five of six readings were averaged while at earlier ages the last two of three readings were averaged and used for the study analyses. Systolic and diastolic (pre)hypertension (SBP-PH/H and DBP-PH/H respectively) were defined using the adult Joint National Committee on the Prevention, Detection, Evaluation, and Treatment of High BP (JNC-7) criteria ( SBP-PH as SBP mmhg; DBP- PH as DBP mmhg and SBP-H as SBP 140 mmhg; DBP-H as DBP 90 mmhg. Participants were weighed with a Wedderburn Chair Scale (nearest 100 g) after fasting and wearing minimal clothing. Height was measured without shoes using a Holtain Stadiometer, Holtain Ltd., Crosswell, UK (nearest 0.1 cm). Body mass index was calculated as weight (kg)/height (m) 2 and was categorized as normal/underweight (<25 kg/m 2 ), overweight (25-30 kg/m 2 ) or obese ( 30 kg/m 2 ) [30]. A priori selected covariates The pre-pregnancy maternal BMI was calculated from self-reported weight before pregnancy and height measured at 18 weeks of gestation. Birth weight (BW) of the offspring was extracted from hospital records. Maternal education was used as a proxy for the socioeconomic status (SES) of the Raine Study family. Mothers with post-school higher education such as 1) trade certificate or apprenticeship 2) professional registration (non-degree) 3) college diploma or degree, or 4) university degree, were ranked as having a higher SES. Stress perceived by the offspring was scored on the shorter version of Depression, Anxiety and Stress Scale (DASS-21) [31] and used as a priori covariate that allowed for offspring current stress. Maternal hypertension comprised hypertensive disorders during pregnancy including 1) gestational hypertension 2) pre-eclampsia 3) pre-pregnancy hypertension 4) pre- 57

72 pregnancy hypertension with superimposed pre-eclampsia. The dichotomous response (yes/no) to questions about offspring smoking, alcohol consumption and female hormonal contraceptive (HC) usage was used in analyses. HC use in young adult females pertained to their current use of any hormonal contraceptive pill, implant, injection or intrauterine device. The short form of the International Physical Activity Questionnaire scoring protocol was used to derive a metabolic equivalents (METS) measure as an index of resting metabolic rate [32]. One metabolic equivalent (MET) is defined as the amount of oxygen consumed while sitting at rest and is equal to 3.5 ml O 2 per kg body weight multiplied by minutes. The score was presented as MET-hours per week, estimated for each individual multiplying by a weighing factor of 3.3, 4 and 8 for walking, moderateintensity and vigorous-intensity activity respectively [32]. Statistical analyses Summary estimates for continuous and categorical variables were calculated as means or proportions with their 95% confidence limits (CI). Differences in characteristics were investigated using 95% CIs for multiple reasons they reflect sample size in their width and give a range of values in which the population value is expected to lie. They also allow differences to be detected when the CIs do not overlap. In order to investigate the relative influence of the stressors, an exploratory factor analysis was performed on the tetrachoric correlations of their binary responses. Cross-sectional analyses Least squares regression was employed to analyse continuous SBP, DBP and BMI outcome models. Both standardized and unstandardized coefficients were calculated for linear regression. Standardized regression coefficients, expressing the change in standard deviations of the outcome resulting from a standard deviation change in the covariate, allowed comparison of effect sizes across variables within a model. [ Logistic regression was used for dichotomous (pre)hypertension and overweight (obesity) outcome models. An interaction term of pregnancy life stress events and BMI investigated the modifying effect of latter on the relationship of maternal life stress and BP, at age 20. Importantly for the primary hypothesis, the cross sectional analysis allowed adjustment for age 20 covariates such as (current stress, smoking, alcohol and 58

73 female HC use) as well as maternal covariates during pregnancy (hypertension, education, and smoking) and offspring covariates at birth (birthweight). These covariates were common to all multivariable linear and logistic regression crosssectional models. In the BMI outcome model, maternal BMI was a covariate in the model. Cohort follow-up yielded additional study variables such as offspring breast feeding at follow-up year 1, familial hypertension at year 8, maternal Depression, Anxiety, Stress scores at 10, 14 and 17 years and young adult physical activity (METS) at year 20. The significance of their inclusion in the final SBP, DBP and BMI cross-sectional models was separately tested. Likelihood ratio tested difference between the nested model pairs, with and without each individual term. A p-value of <0.05 was considered to be statistically significant. Life stress events effect on offspring BP was tested for a higher number of prenatal life stresses on the same set of multivariable regression models. A dichotomous variable for 3 life stress events represented the top tertile of the cumulative life stress events score experienced by the mothers of the study participants. Robust standard errors were generated from the clustered analyses to adjust for correlations between 2 siblings in 20 families. Longitudinal analyses In order to investigate whether similar life stress event effects were observed in earlier years, hierarchical linear mixed models with random slopes using maximum likelihood estimation were employed to analyse the effect of life stress events on the repeated SBP and BMI measures. Combinations of fractional polynomial terms were investigated and the best fit was identified based on the lowest Akaike Information Criteria (AIC). Regression slope estimates for life stress events at each year were derived by postregression contrasts. Analysis was performed using STATA 12 (StataCorp, Texas, USA). 59

74 3.5 Results Life stress events related questions were answered by 2549 women at both 18 and 34 weeks of gestation (Figure 3.1). Figure 3.1. Flow diagram of Raine Study participants attending the 20-year follow-up. At 20 years of age, 1348 Raine study participants attended the follow-up. The final analyses were performed on 957 participants and their mothers that had available all major phenotype and a priori selected study covariates. A comparison with those excluded shows more mothers of excluded participants smoked during the first trimester of pregnancy (P<.001), had lower post-school higher education (P<.001), consumed less alcohol (P<.001), were younger (P<.001) and experienced more life stress events during pregnancy (P=0.001) (Table 3.1). 60

75 Table 3.1. Comparative characteristics of mothers of Raine Study participants versus mothers of non-participants in the 20-year follow-up. Total mothers Mothers Mothers P-value who recorded (n=957) of (n=1592) of for CHARACTERISTICS their life stress participants in excluded difference events (N=2549) the 20-year participants (Col.2 vs (Col.1) analysis (Col.2) (Col.3) 3) Prenatal life stress events 1.5 (1.4, 1.5) 1.4 (1.3, 1.4) 1.5 (1.5, 1.6) Mother age (years) 27.8 (27.6, 28.0) 29.1 (28.8, 29.5) 27.0 (26.7, 27.3) <.001 Pre-pregnancy BMI (kg/m 2 ) 22.4 (22.2, 22.5) 22.3 (22.0, 22.5) 22.4 (22.2, 22.7) Maternal post-school higher education (%Yes) 49.2 (47.2, 51.1) 57.7 (54.6, 60.8) 44.0 (41.6, 46.5) <.001 Maternal smoking in pregnancy (%Yes)* 26.1 (24.5, 27.9) 18.9 (16.4, 21.4) 30.5 (28.2, 32.7) <.001 Maternal alcohol consumption (%Yes)* 46.0 (44.0, 47.9) 50.5 (47.3, 53.6) 43.3 (40.8, 45.7) <.001 Maternal history of hypertension (%Yes) 26.0 (24.3, 27.8) 24.9 (22.1, 27.6) 26.8 (24.6, 28.9) Gestational age (weeks) 39.0 (38.9, 39.0) 39.0 (38.9, 39.1) 39.0 (38.9, 39.0) Offspring birth weight (kg) 3.4 (3.3, 3.4) 3.4 (3.3, 3.4) 3.4 (3.3, 3.4) NB: Data are presented as mean (95% CI) or proportion (95% CI) *Smoking and drinking is related to first trimester of pregnancy 61

76 Cohort characteristics The Raine Study mothers were pregnant at an average age of 29.1 (95%CI: 28.8, 29.5) years with a pre-pregnancy BMI of 22.3 (95%CI: 22.0, 22.5) kg/m 2, 57.7% (95%CI: 54.5, 60.8) had tertiary education and 18.9% (95%CI: 16.4, 21.4) had smoked during the first trimester of their pregnancy (Table 3.2). Among the cohort of 957, 20 offspring were siblings. Table 3.2. Prenatal characteristics of the Raine Study participants mothers. Prenatal characteristics x or % ( 95% CI) Prenatal life stress events 1.4 (1.3, 1.4) Prenatal life stress events 3 (%) 17.2 (15.0, 19.8) Maternal age (years) 29.1 (28.8, 29.5) Pre-pregnancy BMI (kg/m 2 ) 22.3 (22.0, 22.5) Post-school higher education (%) 57.7 (54.5, 60.8) Smoking in 1 st trimester of pregnancy (%) 18.9 (16.4, 21.4) Pregnancy problems (%) 38.3 (35.3, 41.4) Death of a close relative/close friend (%) 11.7 (9.7, 13.7) Death of a close relative (%) 8.5 (6.7, 10.2) Death of a close friend (%) 3.9 (2.6, 5.1) Separation or divorce/marital problems (%) 14.1 (11.9, 16.3) Separation or divorce (%) 4.2 (2.9, 5.4) Marital problems (%) 11.9 (9.9, 14.0) Problems with own children (%) 8.7 (6.9, 10.5) Own job loss/partner job loss/money problems (%) 35.8 (32.8, 38.9) Own job loss (not voluntary) (%) 3.7 (2.5, 4.8) Partner's job loss (not voluntary) (%) 7.8 (6.1, 9.5) Money problems (%) 32.9 (29.9, 35.9) Residential move (%) 26.9 (24.0, 29.7) NB: Data are presented as mean (95% CI) or proportion (95% CI) 62

77 Maternal experience of pregnancy life stress events Mothers experienced an average of 1.4 life stress events during pregnancy and 17.2% (95%CI: 15.0, 19.8) reported a cumulative exposure score of >3 prenatal life stress events (Table 3.2). The three events most commonly reported related to medical or psychological problems that linked to pregnancy, financial, and residence-moving (38.3%, 35.8% and 26.9% respectively) while relationship issues, death of a close relative or a friend, and problems with children were less frequent (14.1%, 11.7% and 8.7% respectively). Factor analysis of life stress items yielded a single factor with loadings suggesting that financial stressors weighted most heavily followed by marital life stresses. Table 3.3. Characteristics of the Raine young adult study participants stratified by gender. CHARACTERISTICS Total (N=957) Males (n=457) Females (n=500) Systolic blood pressure (mmhg) (115.9, 117.4) (121.5, 123.6) (110.4, 112.2) Systolic blood pressure 120 (mmhg) (%) Prehypertension (>=120 to <140 mmhg) (%) 37.0 (33.9, 40.1) 57.8 (53.2, 62.3) 18.0 (14.6, 21.4) 33.4 (30.5, 36.5) 51.0 (46.4, 55.6) 17.0 (14.3, 21.0) Hypertension (>=140 mmhg) (%) 3.6 (2.5, 4.9) 6.8 (4.8, 9.5) 0.3 (0.2, 1.9) Diastolic blood pressure (mmhg) 65.5 (65.0, 66.0) 65.4 (64.7, 66.1) 65.6 (64.9, 66.2) Diastolic blood pressure 80 (mmhg) (%) 3.1 (2.0, 4.2) 3.3 (1.6, 4.9) 3.0 (1.5, 4.5) Birth weight (kg) 3.4 (3.3, 3.4) 3.4 (3.3, 3.4) 3.3 (3.3, 3.4) Young Adult BMI (kg/m 2 ) 24.5 (24.1, 24.8) 24.4 (24.0, 24.8) 24.6 (24.0, 25.1) BMI < 25 (%) 66.5 (63.5, 69.5) 65.6 (61.3, 70.0) 67.2 (63.1, 71.3) BMI 25 and <30 (%) 20.8 (18.2, 23.4) 23.2 (19.3, 27.1) 18.6 (15.2, 22.0) BMI 30 (%) 12.7 (10.6, 14.9) 11.2 (8.3, 14.1) 14.2 (11.1, 17.3) Young Adult stress score 8.9 (8.4, 9.4) 7.3 (6.6, 7.9) 10.4 (9.7, 11.2) Young Adult smoking (%) 14.0 (11.8, 16.2) 15.1 (11.8, 18.4) 13.0 (10.1, 16.0) Young Adult drinking alcohol (%) 68.8 (65.8, 71.7) 72.0 (67.8, 76.1) 65.8 (61.6, 70.0) Physical activity (METS-hrs/week) 59.8 (55.9, 63.8) 78.8 (71.8, 85.8) 42.6 (39.0, 46.2) Maternal hypertension history (%) 24.9 (22.1, 27.6) 28.4 (24.3, 32.6) 21.6 (18.0, 25.2) NB: Data are presented as mean or proportion (95% CI) Standard Error adjusted for 20 families who had 2 children in the study 63

78 BP of offspring at 20 years of age Mean SBP was (95%CI: 115.9, 117.4) mmhg with males having on average 11.3 mmhg higher SBP than females. The prevalence of SBP 120 mmhg was 37%, higher in males than females (57.8% vs 18%). In males and females combined, 33.4% had SBP-PH (SBP mmhg) and 3.6% had SBP-H (SBP 140 mmhg), with a significant gender difference (Table 3.3). At 20 years of age mean DBP was 65.5 mmhg with no significant gender difference (Table 3.3). Diastolic PH/H (DBP 80 mmhg) was evident in 3.1% (30 participants) of whom 83.3% had concomitant SBP- PH/H, with no significant gender difference. Prenatal stress exposure and BP at 20 years Prenatal life stress events were inversely related to offspring SBP (P=0.013) and DBP (P=0.047) (Table 3.4). There was a significant interaction between life stress events score and BMI at 20 years on SBP (P=0.009) indicating that a higher BMI associated with a steeper slope for the inverse relationship between life stress events score on SBP at 20 years (Figure 3.2). Each additional maternal life stress event associated with a 0.66 mmhg reduction in SBP in offspring of average BMI. Table 3.4. Multivariable linear regression outcomes of the association between exposure to prenatal life stress events and BP at 20 years. (N=957) SBP (Model with an interaction) # DBP (No interaction) # Coefficient (95% CI) P value Coef. (95% CI) P value Prenatal life stress events (PregLSE) (-1.19, -0.14) (-0.79, -0.01) Young Adult BMI 0.99 (0.77, 1.20) < (0.05, 0.24) Interaction of PregLSE and BMI ţ (-0.22, -0.03) Female using hormonal contraceptive* 2.22 (0.52, 3.93) (-0.32, 2.22) Male* (10.69, 14.06) < (-0.95, 1.51) Young Adult drinking alcohol 1.47 (0.09, 2.86) (-0.54, 1.51) Young Adult smoking 1.22 (-0.80, 3.24) (-0.46, 2.44) Young Adult stress score (-0.16, 0.004) (-0.08, 0.04) Birth weight (kg) (-3.47, -0.78) (-1.58, -0.23) Maternal hypertension 2.67 (1.17, 4.16) < (0.27, 2.43) Maternal post-school higher education ( -1.96, 0.72) (-1.93, -0.01) Maternal smoking in pregnancy ( -1.87, 1.65) (-1.49, 1.00) Model Constant (112.2, 122.5) < (60.2, 68.4) <.001 #SE adjusted for 20 families having siblings *Female not using HC is the reference level ţbmi is mean-centered in SBP outcome model that allows interpretation of prenatal life stress event-score regression coefficient based on average BMI. 64

79 Figure 3.2. Interaction plot depicting adjusted means of young adult systolic blood pressure for prenatal life stress events score, higher BMI levels accentuating the inverse relationship. Table 3.5. Standardized regression coefficients (beta coefficient) derivation following multivariable linear regression modelling of the association between exposure to prenatal life stress events and BP at 20 years. (N=957) Prenatal life stress score (PregLSE) Y20 SBP model Beta Coefficient # 65 P-value Y20 DBP model Beta Coefficient # P-value Young Adult BMI 0.42 < Interaction of PregLSE and BMI ţ Not Applicable Female using HC* Male* 0.50 < Young Adult drinking alcohol Young Adult smoking Young Adult stress score Child birth weight (kg) Maternal hypertension 0.09 < Maternal tertiary education Maternal prenatal smoking # Regression coefficients with both the x-variable (the independent variable) and the y-variable (the dependent variable) in standard deviations ţbmi is mean-centered in SBP outcome model *Female not using HC is the reference level P-values derived from SE adjusted model for 20 families who had 2 children in the study

80 Maternal history of hypertension during pregnancy (P<.001), being male (P<.001) or female using HC (P=0.011) and drinking alcohol (P=0.037), all related to an increased SBP at 20 years of age (Table 3.4). Higher birth weight had a significant negative effect on SBP (P=0.002). The effect size of life stress events score on offspring SBP, determined by standardized coefficients, was comparable to birth weight and current alcohol consumption (Table 3.5). Table 3.6. Comparative modelling to assess breast feeding, familial BP and maternal DASS score covariates for Y20 SBP and DBP outcomes. OUTCOME Sample size R 2 Root mean 2 PregLSE Coef. (95% CI) Reference: SBP model (at Table 3.4) (-1.19, -0.14) LR test P value Breast feeding(bf) Reference SBP model without BF (-1.33, -0.19) Reference SBP model with BF (-1.32, -0.17) Familial BP (FBP) Reference SBP model without FBP (-1.48, -0.24) Reference SBP model with FBP (-1.49, -0.25) Mean Maternal DASS score (maternal DASS) Reference SBP model without maternal DASS (-1.26, -0.11) Reference SBP model with maternal DASS (-1.22, -0.05) Reference: DBP model (at Table 3.4) (-0.79, -0.01) Breast feeding (BF) Reference DBP model without BF (-0.90, -0.06) Reference DBP model with BF (-0.90, -0.06) Familial BP (FBP) Reference DBP model without FBP (-1.06, -0.15) Reference DBP model with FBP (-1.07, -0.16) Mean Maternal DASS score (maternal DASS) Reference DBP model without maternal DASS (-0.89, -0.06) Reference DBP model with maternal DASS (-0.91, -0.06) Stepwise addition of breast feeding, familial hypertension and maternal combined depression, anxiety, stress score covariates to the final SBP and DBP models revealed no significant difference between the nested models i.e., with and without the individual 66

81 stepwise variables. Likelihood ratio test P values for the nested model differences relating to SBP models were 0.274, 0.071, and relating to DBP models were 0.861, 0.209, and respectively (Table 3.6). Longitudinal analysis for the repeated measures of SBP The slope for prenatal life stress score relation with SBP varied at cohort follow-up years that corresponded to the average offspring age of 3, 5, 8, 10, 14, 17, 20 years. Model fit polynomial terms for the mixed effect longitudinal SBP regression were (i) year (ii) inverse year square (iii) inverse year and (iv) product of inverse year square and log year. A significant (P=0.002) interaction of life stress score and year for repeated SBP measures indicated variation in the life stress score slope over time (Table 3.7). Table 3.7. Longitudinal mixed effect model analysis for repeated SBP outcome measures. SBP outcome Regression coefficient (95% CI) P value Prenatal life stress score (PregLSE) 0.22 (-0.11, 0.55) Year 0.46 (0.26, 0.65) <.001 Interaction - PregLSE and Year (-0.07, -0.02) Male gender* (-17.30, ) <.001 Interaction - Gender and Year 1.25 (1.10, 1.39) <.001 Inverse Year squared ( , ) <.001 Interaction - Gender & Inverse Year squared (-37.50, ) <.001 Inverse Year (141.70, ) <.001 Interaction - Gender & Inverse of Year (30.66, 54.81) <.001 Product of Log Year & Inverse Year squared ( , ) <.001 Young adult BMI (kg/m 2 ) 0.84 (0.77, 0.90) <.001 Maternal hypertension 1.84 (1.23, 2.46) <.001 Maternal smoking during pregnancy 0.17 (-0.48, 0.83) Child birth weight (kg) (-1.84, -0.82) <.001 Model Constant (72.92, 84.64) <.001 *Female is the reference level 67

82 In other words, life stress score relation with SBP was inverse and not significant at the age 8 (-0.14: P=0.228) but marginally significant at age 10 (-0.23: P=0.050). Years 14 and 17 through to young adulthood saw a consistent increase in inverse relationship that was highly significant. The regression coefficients at age 14, 17 and 20 were (P=0.005), (P=0.001) and (P=0.001) (Table 3.8). Table 3.8. Linear mixed model covariate adjusted estimates for SBP and BMI as a function of prenatal life stress events exposure and their interaction with time. SBP outcome BMI outcome Coefficient (95% CI) P value Coefficient (95% CI) P value Year (-0.18, 0.36) (-0.04, 0.06) Year (-0.25, 0.24) (-0.02, 0.10) Year (-0.36, 0.09) (0.02, 0.18) Year (-0.45, ) (0.06, 0.25) Year (-0.61, -0.11) (0.12, 0.36) <.001 Year (-0.80, -0.19) (0.17, 0.48) <.001 Year (-1.06, -0.29) (0.25, 0.64) <.001 Table 3.9. Multivariable logistic regression outcomes of the association between exposure to prenatal life stress events and young adult systolic BP (Prehypertension/Hypertension combined). Odds Ratio P value OUTCOME - Y20 SBP 120 (mmhg) (N=957) (95% CI # ) Prenatal life stress score (PregLSE) 0.83 (0.72, 0.95) Young Adult BMI 1.22 (1.16, 1.29) <.001 Interaction of PregLSE and BMI ţ 0.97 (0.95, 0.99) Female using hormonal contraceptive* 1.35 (0.79, 2.31) Male* 9.82 (6.04, 15.98) <.001 Young Adult drinking alcohol 1.22 (0.86, 1.72) Young Adult smoking 1.33 (0.82, 2.15) Young Adult stress score 0.98 (0.96, 1.00) Child birth weight (kg) 0.54 (0.39, 0.76) <.001 Maternal h/o hypertension 1.64 (1.16, 2.32) Maternal post-school higher education 0.77 (0.56, 1.06) Maternal smoking in pregnancy 0.87 (0.58, 1.31) Model Constant 1.63 (0.42, 6.31) #SE adjusted for 20 families who had 2 children in the study *Female not using HC is the reference ţbmi is mean-centered in the model that allows interpretation of PregLSE regression coefficient based on average BMI 68

83 Prenatal stress exposure and Prehypertension/Hypertension at 20 years Prenatal life stress events were inversely related with prevalence of SBP-PH/H in the offspring at 20 years with an average BMI (P=0.008) (Table 3.9). Each additional life stress event associated with 17% (OR=0.83; 95%CI ) reduction in the offspring s odds for having SBP-PH/H. Maternal history of hypertension (P=0.006) and male gender (P<.001) increased the risk of offspring SBP-PH/H. Higher birth weight (P<.001) or a higher offspring stress score at 20 years (P=0.018) associated with a reduced risk of having SBP-PH/H, independent of having a life stress event. A life stress events score 3 (top tertile) i.e., those most stressed in pregnancy reduced the odds for SBP-PH/H at 20 years by 42% (P=0.016) (Table 3.10). Table Multivariable logistic regression outcomes of the association between exposure to 3 prenatal life stress events and young adult systolic BP (Prehypertension/Hypertension combined). OUTCOME - Y20 SBP 120 (mmhg) (N=957) Odds Ratio (95% CI # ) P value Prenatal life stress events (0.37, 0.90) Young Adult BMI 1.16 (1.13, 1.20) <.001 Female using hormonal contraceptive* 1.34 (0.79, 2.27) Male* 9.50 (5.89, 15.30) <.001 Young Adult drinking alcohol 1.24 (0.88, 1.76) Young Adult smoking 1.30 (0.81, 2.10) Young Adult stress score 0.98 (0.96, 1.00) Child birth weight (kg) 0.56 (0.40, 0.78) Maternal h/o hypertension 1.64 (1.16, 2.32) Maternal post-school higher education 0.78 (0.56, 1.07) Maternal smoking in pregnancy 0.88 (0.58, 1.32) Model Constant 0.03 (0.01, 0.13) >.001 #Standard Error adjusted for 20 families who had 2 children in the study *Female not using HC is the reference level 69

84 Prenatal stress exposure and BMI at 20 years Prenatal life stress events score was positively related to young adult offspring BMI at 20 years (P=0.022) (Table 3.11). Each additional life stress event associated with a 0.37 kg/m 2 increase in offspring BMI at 20 years. Hormonal contraception use among females had a varying effect on BMI. Females using hormonal contraceptives had a negative relation of life stress events score and BMI whereas females not using hormonal contraception, had a significantly higher predicted adjusted mean slope for BMI (Table 3.11). Table Multivariable linear regression outcomes of the association between exposure to prenatal life stress events and BMI at 20 years. OUTCOME - Y20 BMI (kg/m 2 ) (N=957) Coefficient (95% CI # ) P value Prenatal life stress events 0.37 (0.05, 0.68) Maternal pre-pregnancy BMI 0.42 (0.32, 0.52) <.001 Female using hormonal contraceptive* (-2.21, -0.33) Male* (-1.71, 0.77) Young Adult drinking alcohol (-1.47, -0.05) Young Adult smoking (-1.14, 0.72) Young Adult stress score 0.03 (-0.01, 0.07) Birth weight (kg) 0.44 (-0.15, 1.03) Maternal hypertension 0.45 (-0.30, 1.19) Maternal post-school higher education (-0.81, 0.46) Maternal smoking in pregnancy 1.98 (1.04, 2.92) <.001 Model Constant (11.00, 16.64) <.001 #Standard Error adjusted for 20 families who had 2 children in the study *Female not using HC is the reference level 70

85 Adding stepwise breast feeding, maternal depression, anxiety, stress scores and physical activity covariates to the final cross-sectional BMI model, revealed no significant difference among the nested models with and without the individual covariates. Likelihood ratio test derived P values in the nested BMI models were 0.629, and for the respective terms of breast feeding, maternal Depression, Anxiety, Stress scores and physical activity (Table 3.12). Table Comparative modelling to assess breast feeding, maternal DASS and physical activity covariates for Y20 BMI outcome. N R 2 Root mean 2 Prenatal life stress score Coef. (95% CI) LR test P value Reference: BMI model (at Table 3.11) (0.05, 0.68) Breast feeding (BF) Reference BMI model without BF (0.05, 0.59) Reference BMI model with BF (0.05, 0.58) Mean Maternal DASS score (maternal DASS) Reference BMI model without maternal DASS (0.06, 0.60) Reference BMI model with maternal DASS (0.07, 0.61) METS-hours/week Reference BMI model without METS (0.10, 0.62) Reference BMI model with METS (0.10, 0.62) Longitudinal analysis for the repeated measures of BMI The model fit terms for the longitudinal BMI regression were (i) year (ii) year square and (iii) log year. The slope for prenatal life stress events score relation with BMI varied at cohort follow-up years, indicated by a highly significant interaction of life stress events score with time both with fitting terms of year (P<.001) and inverse of year (P=0.012) (Table 3.13). 71

86 Table Longitudinal mixed effect model analysis for repeated BMI outcome measures. BMI outcome Coefficient (95% CI) P value Prenatal life stress score (-0.07, 0.05) Year 1.39 (1.33, 1.45) <.001 Interaction between Prenatal life stress score and Year 0.03 (0.02, 0.05) <.001 Inverse Year (-4.00, -3.66) <.001 Interaction between Prenatal life stress score and Inverse Year (-0.13, -0.02) Male gender* 0.66 (0.50, 0.82) <.001 Interaction between Gender and Year (-0.20, -0.13) <.001 Year squared (-0.02, -0.02) <.001 Interaction between Gender & Year squared 0.01 (0.005, 0.01) <.001 Young adult BMI (kg/m 2 ) 0.04 (0.03, 0.06) <.001 Maternal hypertension 0.14 (-0.004, 0.27) Maternal smoking during pregnancy 0.31 (0.17, 0.45) <.001 Child birth weight (kg) 0.57 (0.46, 0.69) <.001 Model Constant (12.08, 13.04) <0.001 *Female is the reference level The relationship between prenatal life stress events score and BMI was positive and significant from the age 8 (0.10: P=0.010), gradually increasing at age 10 (0.16: P=0.001) through age 14 (0.24: P<.001) and 17 years (0.32: P<.001) to young adulthood at age 20 (0.44: P<.001) (Table 3.8). 72

87 Prenatal stress exposure and young adulthood overweight and obesity Prenatal life stress events score was positively related to offspring overweight and obesity (P=0.039) (Table 3.14)]. Each additional life stress associated with 14% higher odds for overweight/obesity. Table Multivariable logistic regression outcomes of the association between exposure to prenatal life stress events and young adult Overweight/Obesity combined. OUTCOME - Y20 BMI 25 (kg/m 2 ) Odds Ratio (95% CI) P value Prenatal life stress score 1.14 (1.01, 1.30) Maternal pre-pregnancy BMI 1.15 (1.11, 1.19) <.001 Female not using hormonal contraceptive* 1.57 (1.05, 2.35) Male* 1.31 (0.94, 1.84) Young Adult drinking alcohol 0.76 (0.55, 1.03) Young Adult smoking 0.91 (0.59, 1.41) Young Adult stress score 1.01 (0.99, 1.03) Child birth weight (kg) 1.27 (0.95, 1.69) Maternal hypertension 1.38 (0.99, 1.92) Maternal post school higher education 1.01 (0.74, 1.36) Maternal smoking in pregnancy 1.96 (1.35, 2.85) <.001 Model Constant 0.01 (0.002, 0.02) <.001 NB: #Standard Error adjusted for 20 families who had 2 children in the study *Female using HC is the reference level Maternal pre-pregnancy BMI (P<.001) and maternal smoking in pregnancy (P<.001) increased the risk of overweight/obesity. HC non-user females had an increased risk of 1.57 times relative to females on HC for being overweight/obese (P=0.028) (Table 3.14). 73

88 3.6 Discussion In contrast to our prediction this study has shown that greater exposure to life stress during pregnancy and in particular life stress related to financial problems, associated with lower SBP and 17% lower risk for the development of systolic (pre)hypertension in offspring at 20 years. This was despite demonstrating a positive association between maternal pregnancy life stress scores and offspring BMI. Higher levels of BMI accentuated the inverse relationship between prenatal life stress and offspring BP. The inverse effects of maternal life stress events on SBP were clearly evident at age 14 and later while the effect of maternal life stress events on higher BMI trajectory started as early as age 8. We are not aware of comparable studies that report associations between pregnancy life stress scores and early adult BP. Only one of two similar Dutch childhood cohort studies found a significant positive association between early life stress and 6-year-old offspring BP [19, 20]. In both these studies, psychosocial predictor information was only collected in early pregnancy. More recently, the UK-based The Avon Longitudinal Study of Children and Parents (ALSPAC) longitudinal cohort study reported that maternal psychosocial stress (depression/anxiety) during pregnancy associated with lower DBP but not systolic pressure in 10-12y offspring (p<0.05) [21]. In our study, prenatal stress showed a negative relation with DBP (P=0.047). Importantly, our study demonstrates a significant dose response inverse relationship between prenatal stress scores and early adulthood, independent of measures of current stress. The finding of an inverse association between maternal pregnancy life stresses and adult BP accentuated by a higher BMI was unexpected: if causal, the mechanisms are unclear. It is possible that the effects are due to unrecognised confounders or underestimates of the effects of known confounders. However, a dose response effect and the impact of adiposity suggest they may be causal. It is possible that maternal pregnancy life stresses are a marker for continuing postnatal family environment that predisposes to low blood pressure. If the inverse association between maternal life stress in pregnancy and adult offspring BP is causal, then possible mechanisms may relate to effects of foetal programming on BP regulation via epigenetic mechanisms [33]. The hypothalamic pituitary adrenal (HPA) axis has been a particular focus of animal studies [34]. Foetal overexposure to 74

89 maternal stress hormones is normally safeguarded by placental 11β-hydroxysteroid dehydrogenase (11β-HSD), which can inactivate cortisol and prevent its passage to the foetal blood circulation. Dysregulation of this key component of the feto-placental glucocorticoid barrier is thought to overexpose the foetus to stress-induced increased levels of maternal glucocorticoids [35]. Consequent epigenetic changes within the foetal HPA-axis have been proposed to compromise adrenal angiotensin II type 1b receptor and hypothalamic glucocorticoids receptor function [2]. Such an indirect influence of maternal stress on the set point of the foetal HPA and adrenomedullary stress-regulatory systems, may contribute to altered physiological responses to stress in later life [36-42]. However, the animal data suggests that various stressors in pregnancy are associated with higher rather than lower BP in offspring resulting in a hypertensive response to stress-induced enhanced cortisol output [43-46]. At the same time, an adaptive response to foetal programming may maintain or slightly down regulate BP-maintenance under stress-free conditions [47]. For example, young adult Wistar rats prenatally treated with the synthetic glucocorticoid dexamethasone displayed a lower basal BP in adulthood, despite a significantly greater rise in BP in response to stress [47]. In our study, the putative lowering effect of maternally perceived prenatal life stress on resting SBP and SBP-PH/H could be considered akin to a lower basal BP finding of the latter animal study. Our study did not include all Raine Study participants who took part in the 20-year follow up and had missing values for few covariates collected between periods of infancy and young adulthood. However, as the original cohort was recruited predominantly from public hospitals, the Raine Study cohort is now more representative of the Western Australian population [28], with the greater retention of socially advantaged families over time [29]. It was not possible to have a formal clinical psychological evaluation of the participant mothers to diagnose depressive status. In this study, maternal depression, anxiety and stress scores were obtained through questionnaire which is well established and practical procedure to ascertain negative emotional status of individuals in large population studies. Strengths of this study include prospective assessment of effects of exposure to major life stress events in utero on population based young adult offspring BP and BMI in a well characterised cohort with carefully documented major confounders. These findings were strengthened by longitudinal modelling of data obtained throughout childhood 75

90 which supported the cross-sectional findings at year 20. This study has shown a dose response for the inverse relationships between prenatal life stress, particularly financial problems, and BP in offspring at 20 years of age, independent of current offspring stress. Maternal pregnancy life stress associates positively with offspring BMI at 20 years, the significant relation starting in early childhood. Although this suggests a high degree of plausibility for a causal effect, we cannot exclude influences of unidentified confounders or that life stress events during infancy and childhood could influence offspring health. In conclusion, we have demonstrated that despite pregnancy maternal life stress events being associated with an increase in offspring BMI such stress paradoxically corresponded with lower BP and less risk of (pre)hypertension. This phenomenon clearly warrants replication in other populations and further evaluation as to the possible underlying mechanisms. 76

91 3.7 References [1]. Barker DJP, Osmond C, Forsen TJ, Kajantie E, Eriksson JG. Maternal and social origins of hypertension. Hypertension. 2007;50: [2]. Gluckman PD, Hanson MA, Cooper C, Thornburg KL. Effect of in utero and early-life conditions on adult health and disease. New England Journal of Medicine. 2008;359:61-73 [3]. Ravelli G-P, Stein ZA, Susser MW. Obesity in young men after famine exposure in utero and early infancy. New England Journal of Medicine. 1976;295: [4]. Ó Gráda C. Fetal origins, childhood development, and famine: A bibliography and literature review. Working Paper Series. UCD School of Economics, University College Dublin (UCD) Accessed May 1, 2014 [5]. Lumey L, Stein AD, Susser E. Prenatal famine and adult health. Annual Review of Public Health. 2011;32: [6]. Entringer S. Impact of stress and stress physiology during pregnancy on child metabolic function and obesity risk. Current Opinion in Clinical Nutrition and Metabolic Care. 2013;16: [7]. Flory JD, Bierer LM, Yehuda R. Maternal exposure to the holocaust and health complaints in offspring. Disease Markers. 2011;30: [8]. King S, Barr R, Brunet A, Saucier J, Meaney M, Woo S, Chanson C. The ice storm: An opportunity to study the effects of prenatal stress on the baby and the mother. Santé mentale au Québec. 1999;25: [9]. Li J, Olsen J, Vestergaard M, Obel C, Baker JL, Sørensen TI. Prenatal stress exposure related to maternal bereavement and risk of childhood overweight. PLoS One. 2010;5:e11896; doi: /journal.pone [10]. Li J, Olsen J, Vestergaard M, Obel C, Kristensen JK, Virk J. Prenatal exposure to bereavement and type-2 diabetes: a Danish longitudinal population based study. PLoS One. 2012;8:1-4; doi: /journal.pone

92 [11]. Li M, Sloboda D, Vickers M. Maternal obesity and developmental programming of metabolic disorders in offspring: evidence from animal models. Experimental Diabetes Research. 2011;Vol 2011:1-9; doi: /2011/ [12]. Remacle C, Bieswal F, Reusens B. Programming of obesity and cardiovascular disease. International Journal of Obesity. 2004;28:S46-S53 [13]. Tamashiro KLK, Terrillion CE, Hyun J, Koenig JI, Moran TH. Prenatal stress or high-fat diet increases susceptibility to diet-induced obesity in rat offspring. Diabetes. 2009;58: [14]. Barker DJP. The developmental origins of adult disease. Journal of the American College of Nutrition. 2004;23:588S-595S [15]. Barker DJ, Osmond C, Golding J, Kuh D, Wadsworth ME. Growth in utero, blood pressure in childhood and adult life, and mortality from cardiovascular disease. BMJ. 1989;298: [16]. Godfrey KM, Barker DJ. Fetal programming and adult health. Public Health Nutrition. 2001;4: [17]. Launer LJ, Hofman A, Grobbee DE. Relation between birth weight and blood pressure: Longitudinal study of infants and children. British Medical Journal. 1993;307: [18]. Law CM, Shiell AW. Is blood pressure inversely related to birth weight? The strength of evidence from a systematic review of the literature. Journal of Hypertension. 1996;14: [19]. Van Dijk AE, Van Eijsden M, Stronks K, Gemke RJBJ, Vrijkotte TGM, Jaddoe VWV. The association between prenatal psychosocial stress and blood pressure in the child at age 5-7 years. PloS one. 2012; 7:e43548; doi: /journal.pone [20]. Taal HR, de Jonge LL, Tiemeier H, van Osch-Gevers L, Hofman A, Verhulst FC, Helbing WA, van der Heijden AJ, Jaddoe VWV. Parental psychological distress during pregnancy and childhood cardiovascular development. The Generation R study. Early Human Development. 2013; 89: ; Accessed April 12,

93 [21]. van Dijk AE, Dawe K, Deanfield J, Stronks K, Gemke RJ, Vrijkotte TG, Lawlor DA. The association of maternal prenatal psychosocial stress with vascular function in the child at age years: findings from the Avon longitudinal study of parents and children. European Journal of Preventive Cardiology. 2014; 21: ; doi: / [22]. Uiterwaal CSPM, Anthony S, Launer LJ, Witteman JCM, Trouwborst AMW, Hofman A, Grobbee DE. Birth weight, growth, and blood pressure: An annual followup study of children aged 5 through 21 years. Hypertension. 1997;30: [23]. Barker DJP, Bagby SP, Hanson MA. Mechanisms of disease: in utero programming in the pathogenesis of hypertension. Nature Clinical Practice Nephrology. 2006;2: [24]. Cleasby ME, Kelly PA, Walker BR, Seckl JR. Programming of rat muscle and fat metabolism by in utero overexposure to glucocorticoids. Endocrinology. 2003;144: [25]. Phillips D. Programming of the stress response: A fundamental mechanism underlying the long term effects of the fetal environment? Journal of Internal Medicine. 2007;261: [26]. Entringer S, Buss C, Swanson JM, Cooper DM, Wing DA, Waffarn F, Wadhwa PD. Fetal programming of body composition, obesity, and metabolic function: the role of intrauterine stress and stress biology. Journal of Nutrition and Metabolism. 2012; 2012:632548; doi: /2012/ [27]. Spencer SJ, Tilbrook A. The glucocorticoid contribution to obesity. Stress. 2011;14: [28]. Newnham JP, Evans SF, Michael CA, Stanley FJ, Landau LI. Effects of frequent ultrasound during pregnancy: a randomised controlled trial. The Lancet. 1993;342:

94 [29]. Robinson M, Mattes E, Oddy WH, Pennell CE, van Eekelen A, McLean NJ, Jacoby P, Li J, De Klerk NH, Zubrick SR, Stanley FJ, Newnham JP. Prenatal stress and risk of behavioral morbidity from age 2 to 14 years: the influence of the number, type, and timing of stressful life events. Development and Psychopathology. 2011;23: [30]. WHO Consultation. Obesity: preventing and managing the global epidemic. World Health Organization Technical Report Series. 2000;894:9 [31]. Lovibond SH, Lovibond PF. (1995) Manual for the Depression Anxiety Stress Scales. (2nd Ed) Sydney: Psychology Foundation. [32]. Jurakić D, Pedišić Ž, Andrijašević M. Physical activity of Croatian population: cross-sectional study using international physical activity questionnaire. Croatian Medical Journal. 2009;50: [33]. Ordovas JM, Smith CE. Epigenetics and Cardiovascular disease. Nature Reviews Cardiology. 2010;7: [34]. Seckl JR. Glucocorticoid programming of the fetus; adult phenotypes and molecular mechanisms. Molecular and Cellular Endocrinology. 2001;185:61-71 [35]. Mcmillen IC, Robinson JS. Developmental origins of the metabolic syndrome: prediction, plasticity, and programming. Physiological Reviews. 2005;85: [36]. Goldstein D. Adrenal responses to stress. Cellular and Molecular Neurobiology. 2010;30: [37]. Kapoor A, Dunn E, Kostaki A, Andrews MH, Matthews SG. Fetal programming of hypothalamo-pituitary-adrenal function: prenatal stress and glucocorticoids. The Journal of Physiology. 2006;572:31-44 [38]. Harris A, Seckl J. Glucocorticoids, prenatal stress and the programming of disease. Hormones and Behavior. 2011;59: [39]. Benediktsson R, Lindsay RS, Noble J, Seckl JR, Edwards CRW. Glucocorticoid exposure in utero: new model for adult hypertension. The Lancet. 1993;341:

95 [40]. Van den Bergh BRH, Mulder EJH, Mennes M, Glover V. Antenatal maternal anxiety and stress and the neurobehavioural development of the fetus and child: links and possible mechanisms. Neuroscience & Biobehavioral Reviews. 2005;29: [41]. Karlén J, Frostell A, Theodorsson E, Faresjö T, Ludvigsson J. Maternal influence on child HPA axis: a prospective study of cortisol levels in hair. Pediatrics. 2013; 132:e ; doi: /peds [42]. Reynolds RM. Glucocorticoid excess and the developmental origins of disease: two decades of testing the hypothesis 2012 Curt Richter award winner. Psychoneuroendocrinology. 2013;38:1-11 [43]. Reynolds RM, Walker BR, Syddall HE, Andrew R, Wood PJ, Whorwood CB, Phillips DIW. Altered control of cortisol secretion in adult men with low birth weight and cardiovascular risk factors. Journal of Clinical Endocrinology & Metabolism. 2001;86: [44]. Miller GE, Chen E, Parker KJ. Psychological stress in childhood and susceptibility to the chronic diseases of aging: moving toward a model of behavioral and biological mechanisms. Psychological Bulletin. 2011;137: [45]. Tang JI, Kenyon CJ, Seckl JR, Nyirenda MJ. Prenatal overexposure to glucocorticoids programs renal 11β-hydroxysteroid dehydrogenase type 2 expression and salt-sensitive hypertension in the rat. Journal of Hypertension. 2011;29: [46]. Kett MM. Arterial pressure in animal models of fetal programming: hypertensive or hyperresponsive? Journal of Hypertension. 2012;30: [47]. O'Regan D, Kenyon CJ, Seckl JR, Holmes MC. Prenatal dexamethasone programmes hypotension, but stress-induced hypertension in adult offspring. Journal of Endocrinology. 2008;196:

96 Chapter 4 The role of maternal and offspring life style behaviours in the relationship between depression and adiposity in young adults 4.1 Preamble A possible psychosocial behavioural pathway mediating the positive relationship between depression and obesity is postulated with a particular focus on maternal smoking status and family income. Maternal prenatal smoking is associated with offspring behavioural disorders. Smoking exposure, both active and passive, is an important component of cardiovascular health promotion programmes. Low family income has also been associated with obesity and depression in children. However, it is not known if the association between depression symptoms and adiposity in young adults varies according to maternal prenatal smoking status or family income. Young adults have more established adult behaviour patterns than children. Additionally, comorbidities and prescription medication usage is lower among young adults than in middle aged or elderly subjects. This chapter therefore examined the underlying influence of adverse psychosocial factors including maternal prenatal smoking and low family income at pregnancy on the relationship between depression symptoms and adiposity in Raine participants at age

97 4.2 Abstract There is an increasing incidence of overweight/obesity and mental health disorders in young populations and the two conditions often coexist. This study investigated the influence of antenatal and postnatal factors that may underlie this association with a focus on the influence of maternal smoking in pregnancy, socio-economic status and gender. The Western Australian Raine-study data from 1056 adults aged 20 years were analysed using multivariable models for associations between offspring depression scores (DASS-21 Depression scale) and body mass index (BMI). We observed a significant positive relationship between offspring depression score and BMI independent of gender and other psychosocial covariates. There was a significant interaction between maternal smoking during pregnancy and depression score (Interaction Coefficient=0.096; 95%CI: 0.006, 0.19, P=0.037), indicating the relationship between depression score and BMI differed according to maternal prenatal smoking status. In offspring of maternal prenatal smokers, a positive association between BMI and depression score (Coefficient=0.133; 95%CI: 0.05, 0.21, P=0.001) equated to 1.1kg/m 2 increase in BMI for every 1 standard deviation (8 units) increase in depression score. Substituting low family income during pregnancy for maternal prenatal smoking in the interaction (Interaction Coefficient=0.091; 95%CI: 0.01, 0.17, P=0.027) showed a positive association between BMI and depression score only among offspring of mothers with a low family income during pregnancy (Coefficient=0.118; 95%CI: 0.06, 0.18, P<0.001). There were no significant differences between genders in these associations. These findings indicate important maternal and socio-economic factors that identify individuals vulnerable to the coexistence of obesity and depression in early adulthood. Keywords: depression; maternal prenatal smoking; obesity 83

98 4.3 Introduction The increasing worldwide incidence of overweight and obesity in children and adolescents [1] has been paralleled by an increase in mental health disorders [2]. Several meta-analyses have shown significant positive associations between obesity and depression or anxiety in the general population [3,4]. However, the nature of this association is likely to be complex, with possible bidirectional causality [5] and common underlying familial and environmental [6], genetic [7] and epigenetic [8,9] influences. There are conflicting results concerning gender differences in the association between depression and obesity. Some studies report a significant positive association between depression and obesity in females but not in males while other studies found no gender discrepancy [3]. Boys with clinical depression in childhood have been reported to have higher adult body mass index (BMI) [10]. In contrast, we have previously reported a positive association between depression/anxiety scores and BMI in 14-year-old girls from the Western Australian Pregnancy Cohort (Raine) Study [11], while another study reported an association between major depression in early adolescence and the development of subsequent obesity in women but not men [12]. Variations in methodology with regard to selected populations and varying definitions of depression or use of depression scores as a continuum may account for some of these discrepancies. Possible pregnancy related factors which may also influence any association between depression scores and adiposity in young adults include maternal smoking in pregnancy and socio-economic status. Maternal smoking in pregnancy has been independently associated with offspring obesity [13,14] and childhood behavioural problems [15-17]. Low socio-economic status has been associated, with higher symptoms of depression, anxiety, stress [18,19], smoking in pregnancy [20,21] and obesity [19,22,23]. As the Raine study participants have now reached early adulthood with more established adult behaviour patterns, we aimed to identify those pregnancy and early life factors, and offspring behaviours that may predispose to any association between adiposity and depressive tendencies in young adults. We hypothesized that individuals whose mothers smoked in pregnancy would be more prone to show a positive association between depressive tendencies and BMI, and that any such effect may also be mediated by socioeconomic status. 84

99 4.4 Material and methods The Raine Study is a prospective longitudinal pregnancy study of 2868 live births from pregnant women enrolled at 18 weeks of gestation. The methodology for the study has been described in detail elsewhere [24]. Briefly, 2900 pregnant women were recruited to the study between 16 and 20 weeks gestation through the public antenatal clinic at King Edward Memorial Hospital (KEMH) in Perth, Western Australia, or surrounding private practices between May 1989 and November The potential for introducing bias by using a tertiary referral center sample was minimized by enrolling women who booked before 18 weeks of gestation, which excluded those referred with complications. Ninety percent of eligible women agreed to participate in the study and written informed consent to participate was obtained at enrolment. The study has ethics approval from the Human Research Ethics Committees at King Edward Memorial Hospital, Princess Margaret Hospital for Children and The University of Western Australia. This analysis uses data from the 20-year follow-up of the offspring, conducted from Informed written consent was obtained from the parents or a primary carer at each follow-up and from the participants at 20 years. Anthropometry Weight was recorded with a Wedderburn Chair Scale (nearest 100g) after fasting and wearing minimal clothing. Height was measured without shoes using a Holtain Stadiometer, Holtain Ltd, Crosswell, UK (nearest 0.1cm). BMI was categorized as normal/underweight (<25kg/m 2 ), overweight ( kg/m 2 ) or obese ( 30kg/m 2 ). DASS Depression-score The Depression, Anxiety, Stress Scale (DASS-21) is the short-form of the DASS-42, a combination of three self-report scales and comprising twenty-one questions [25]. Seven questions in each category measure the negative emotional states respectively of depression, anxiety and stress [25]. The DASS-21 scores were multiplied by 2 as per scoring instructions allowing comparison with DASS-42 normative data [25]. DASS-21 highly correlates to other validated measures of depression and anxiety, indicating a high convergent validity [25-27]. Depression symptoms experienced over the past week, were scored on a four-point severity/frequency scale (did not apply to me at all=0, applied to me to some degree, or some of the time=1, applied to me to a 85

100 considerable degree, or a good part of the time=2, applied to me very much, or most of the time=3). The seven symptoms were I could not seem to experience any positive feeling at all. I felt that I had nothing to look forward to, I felt I was not worth much as a person, I felt down-hearted and blue, I was unable to become enthusiastic about anything, I felt that life was meaningless, and I found it difficult to work up the initiative to do things [25]. Maternal psychosocial covariates The pre-pregnancy BMI of the participants mothers was calculated from self-reported weight before pregnancy and height measured at 18-weeks gestation. If women could not answer this question, their weight at the completion of this questionnaire was used to estimate pre-pregnancy weight. Self-report of weight by women of reproductive age has been found to be valid with 84% of women correctly classified into the correct BMI categories [28]. Maternal self-reported data for 5 levels of weekly alcohol intake were obtained at 18 weeks' gestation. Maternal smoking during pregnancy was self-reported for 6 levels of daily smoking. Given that there were no quantified safe levels of drinking and smoking during pregnancy at the time, maternal smoking and alcohol consumption were dichotomized to a yes/no response. Mothers with post-school higher education were ranked as having a tertiary education. Annual family income at the time of pregnancy was dichotomized, with low income status defined as <$24,000 Australian-Dollars (AUD) during the enrolment period ( ), in accordance with the poverty line at the time the data were collected. Breastfeeding data were collated from questionnaires completed by mothers at the 1-, 2- and 3-year follow-ups and was assessed as the age breastfeeding stopped and the introduction of milk other than breastmilk was introduced [29]. At 18-weeks gestation mothers of the offspring were asked whether they had experienced any of ten life-stress events since becoming pregnant [30]. Family-functioning at adolescent age of 14 was determined using the General Functioning Scale from the McMaster Family Assessment Device [31]. Offspring psychosocial covariates Birth weight of the participants was obtained from hospital records. The self-reported 86

101 measures of current smoking and current alcohol consumption at 20 years of age were dichotomized to a yes/no response. Information about use of hormonal contraceptives (HC) in females was based on self-reported current use of the oral contraceptive (OC) pill, implant, injection or any intrauterine HC-device. The short form of the International Physical Activity Questionnaire was used to derive a measure of metabolic equivalents (METS), a unit of resting metabolic rate [32]. Statistical analyses All the analyses were performed using STATA 12 (StataCorp, Texas, USA). Summary estimates for continuous and categorical variables were calculated as means/proportions with 95% CIs. Significant differences between characteristics of participants versus non-participants, maternal prenatal smokers versus non-smokers, and high versus low family income during pregnancy were determined, using chi-square/z tests. Ordinary least square (OLS) models regressed the BMI outcome on DASS-Depression score and depression scores on medical practitioner diagnosed depression. The linearity of the relation was assessed examining scatterplot overlaid with locally weighted regression smoother. Preliminary univariate analyses involved variables of clinical and biological relevance for BMI and Depression-score outcomes [33]. The offspring covariates in these univariate analyses included gender, gestational age at delivery (weeks), birth weight, breast feeding ( 4 months), current alcohol drinking, current smoking and physical activity (hours/week). Maternal covariates included age, history of hypertension in pregnancy, pre-pregnancy BMI, smoking, alcohol consumption and life-stress score during pregnancy, post-school higher (tertiary) education and low family income at the time of pregnancy. All variables potentially associated (P<0.10) with BMI in univariate analyses were included in the initial multivariable regression model. A backward elimination regression identified the final set of significant (P<0.05) covariates. Covariates with the largest P-value above 0.05 were excluded first and at each step the model was examined for evidence of the excluded covariate s importance from the change in regression coefficient and model variance. The two-way interaction between depression and maternal prenatal smoking or family income at pregnancy was introduced separately to the penultimate multivariable model to test whether maternal smoking or family income 87

102 modified the depression and BMI relationship. We also investigated potential effect modification by gender. All regression models used within family cluster adjustment to correct for correlation between a small numbers of siblings. 4.5 Results At 20 years of age 1348 participants attended the follow-up (Figure 4.1). Figure 4.1. Flow diagram of Raine Study participants attending the 20-year follow-up. Complete data on primary variables of interest BMI, DASS Depression score, maternal smoking in pregnancy and gender were available for 1056 participants and their mothers. A comparison of study participants (n=1056) with those that did not participate in the 20-year follow-up (n=1743) shows more mothers of those that did not participate smoked during the first trimester of pregnancy (31.4 vs 19.5%, P<0.001), were less likely to have a tertiary education (43.4 vs 56.6%, P<0.001) or to consume alcohol during pregnancy (42.9 vs 49.9%, P<0.001), were younger (26.8 vs 29.0 years, P<0.001), and were more likely to have a low family income at the time of pregnancy (49.6 vs 34.4%, P<0.001) (Table 4.1). These findings are indicative of some retention bias for families, more likely to be well-educated, health conscious and having a relatively higher income. 88

103 Young adult offspring that did not complete the DASS Depression score questionnaire (n=244) were more likely to have mothers with a low family income at the time of pregnancy (43.2 vs 34.3%; P=0.012) and were more likely to be males (70.1 vs 47.7%; P<0.001) (not shown). Table 4.1. Pregnancy related characteristics for mothers of participants versus mothers of non-participants in the 20-year follow-up. Mothers of participants that did not participate (n=1743) Mothers of participants included (n=1056) P value Mean / Percent 95% CI Mean / Percent 95% CI Age (yr.) , , 29.3 <.001 Pre-pregnancy BMI (kg/m 2 ) , , Pre-pregnancy Obesity (%) , , Hypertension in pregnancy (%) , , Smoking in pregnancy (%) , , 21.9 <.001 Tertiary education (%) , , 59.6 <.001 Family income <$24,000 AUD (%) , , 37.3 <.001 Alcohol drinker in pregnancy (%) , , 52.9 <.001 Gestational age (weeks) , , Offspring birth weight (kg) , , Maternal and Offspring Cohort Characteristics Table 4.2 shows the characteristics of offspring according to gender. Males had lower average depression, anxiety, and stress scores, and a higher level of physical activity compared with females. The proportion of HC-use by females was 60.8%. Depression diagnosed by a medical practitioner was reported in 16.1% of young adult participants. In these participants, depression scores were 8.25 units (95%CI: 7.10, 9.41, P<0.001) higher compared with the scores of participants not clinically diagnosed with depression (not shown). 89

104 Table 4.2. Participants characteristics by gender. Offspring characteristics at 20Y Total (n=1056) Females (n=554) Males (n=502) Mean / Percent 95% CI Mean / Percent 95% CI Mean / Percent 95% CI BMI (kg/m 2 ) , , , 24.8 BMI < 25 (%) , , , 69.6 BMI 25 and <30 (%) , , , 27.4 BMI 30 (%) , , , 14.0 Depression core Ɨ , , , 6.7 Anxiety score Ɨ , , , 4.7 Stress score Ɨ , , , 8.0 Total DASS score , , , 19.2 Hormonal contraceptive use(%) , Alcohol drinker (%) , , , 76.8 Smokers (%) , , , 19.2 Physical activity (METS-hrs/week) , , , 84.5 Birth weight (kg) , , , 3.4 Breast fed (%) , , , 94.7 Breast fed 4 months (%) , , , 73.0 Maternal characteristics Age (yr.) , , , 29.4 Pre-pregnancy BMI (kg/m 2 ) , , , 22.7 Pre-pregnancy Obesity (%) , , , 8.9 Hypertension in pregnancy (%) , , , 32.0 Smoking in pregnancy (%) , , , 20.3 Tertiary education (%) , , , 62.0 Pregnancy life stress score , , , 1.4 Low income in pregnancy (%) , , , 37.9 Family functioning score , , ,

105 Table 4.3. Univariate and Multivariable adjusted BMI regression. N Univariate models Multivariable model* (Significant covariates only) Coef. 95% CI P Coef. 95% CI P Offspring covariates Depression-score , 0.14 < , Gender (Female reference) 1056 Male , Tripartite Gender (No-HCuser reference) 1056 HC-user female , , Male , , Alcohol drinker , Smokers , Physical activity (METShours/week) Gestational age at delivery (weeks) , , Birth weight (kg) , Any breast feeding , Breast feeding 4 months , Maternal covariates Maternal age (yr.) at pregnancy Pre-pregnancy BMI (kg/m 2 ) , , , 0.56 < , 0.54 <.001 Hypertension in pregnancy , Maternal smoking in pregnancy , 2.96 < , 2.57 <.001 Maternal tertiary education in pregnancy , Low family income at pregnancy , Family functioning score at offspring age , Alcohol intake in pregnancy , Life stress events score in pregnancy , 0.84 <.001 *Model constant for multivariable model (regression coefficient=16.59; 95%CI 14.05, 19.13; P<.001) 91

106 Univariate and Multivariable adjusted BMI Regression In univariate analysis, DASS Depression-score associated positively with offspring BMI (0.09; 95%CI: 0.04, 0.14, P<0.001) (Table 4.3). Covariates that significantly associated positively with offspring BMI were birth weight, high life stress score during pregnancy, higher maternal pre-pregnancy BMI, history of maternal hypertension in pregnancy, maternal smoking in pregnancy, and low family income at the time of pregnancy (Table 4.3). Covariates significantly inversely associated with offspring BMI were breast feeding 4 months, drinking alcohol at age 20, higher maternal age, and maternal tertiary education in pregnancy (Table 4.3). In a multivariable model the depression-score and BMI association remained significant after adjusting for gender and HC use in females, maternal pre-pregnancy BMI, maternal age, and maternal prenatal smoking (Table 4.3). Adjustment attenuated the magnitude of the depression-score effect on BMI (0.06; 95%CI: 0.02, 0.10, P=0.002) (Table 4.3). This equated to 0.5 kg/m 2 increase in BMI for every 1 SD (8 units) increase in depression score. Effect of Maternal Prenatal Smoking on the Association Between Depression Scores and BMI The final multivariable regression model showed a significant interaction between maternal prenatal smoking and depression-score (Interaction Coefficient=0.096; 95%CI: 0.006, 0.19, P=0.037), indicating the relationship between depression scores and BMI differed according to the maternal prenatal smoking status (Table 4.4). A positive association between BMI and depression score was found among offspring of maternal prenatal smokers (0.133; 95% CI 0.05, 0.21, P=0.001) that equated to 1.1kg/m 2 increase in BMI for every 1 SD (8 units) increase in depression score. However, no significant association was detected for offspring of mothers that did not smoke in pregnancy (0.037; -0.01, 0.08, P=0.108) (Table 4.4 and Figure 4.2). The relationship in offspring of maternal prenatal smokers was independent of gender and HC-use in females, maternal age, and pre-pregnancy BMI. 92

107 Table 4.4. Multivariable adjusted BMI regression (Final Model). Model Sample (N=1056) Slope Coefficient 95% CI P-value Depression-score ǂ , Maternal smoking (during pregnancy)^ , Interaction Depression-score and Maternal smoking , Adjusted for the following variables HC-user female Ɨ , Male Ɨ , Maternal age (yr.) at pregnancy , Pre-pregnancy BMI (kg/m 2 ) , 0.54 <.001 Model Constant , <.001 ǂ Depression-score coefficient in offspring of maternal prenatal non-smokers Depression-score coefficient in offspring of maternal prenatal smokers = ( ) = 0.133; 95%CI: 0.05, 0.21; P=0.001 (independent of gender, female HC-use, maternal age and prepregnancy BMI) ^Difference in BMI between offspring of maternal smoker vs non-smoker (during pregnancy) for those with depression-score = 0 Ɨ Female not using hormonal contraceptive is the reference Figure 4.2. Interaction plot depicting adjusted means of young adult BMI for maternal prenatal smoking levels, maternal prenatal smoking showing a significant positive association between depression score and BMI compared to depression-score slope for offspring of mothers that did not smoke during pregnancy. 93

108 Table 4.5. Characteristics by maternal smoking or low family income during pregnancy. Offspring characteristics at 20Y Maternal nonsmokers during pregnancy (n=850) µ/ % 95% CI Maternal smokers during pregnancy (n=206) µ/ % 95% CI P value Mothers with high family income during pregnancy (n=673) µ/ 95% CI % Mothers with low family income during pregnancy (n=351) µ/ 95% CI % Female (%) , , , , P value Non-HC user female (%) , , , , HC-user female (%) , , , , Male (%) , , , , BMI (kg/m 2 ) , , 27.0 < , , 25.8 <.001 BMI < 25 (%) , , 62.6 < , , 64.4 <.001 BMI 25 and <30 (%) , , , , BMI 30 (%) , , 25.9 < , , 22.6 <.001 Depression score Ɨ , , , , Anxiety score Ɨ , , , , Stress score Ɨ , , , , Total DASS score , , , , Alcohol drinker (%) , , , , 67.6 <.001 Smokers (%) , , 30.1 < , , Physical activity (METShr/week) , , , , Birth weight (kg) , , 3.3 < , , Any breast feeding (%) , , , , Breast fed 4 months (%) , , 54.5 < , , 61.9 <.001 Maternal characteristics Age (yr.) , , 27.6 < , , 27.1 <.001 Pre-pregnancy BMI (kg/m 2 ) , , , , Pre-pregnancy Obesity (%) , , , , 12.4 <.001 Hypertension in pregnancy (%) , , , , Tertiary education (%) , , 41.0 < , , 45.3 <.001 Pregnancy life stress score , , 1.8 < , , 1.8 <.001 Low income at pregnancy (%) , , 57.0 < Maternal prenatal smoking (%) , , 32.6 <.001 Family functioning score , , , , Ɨ Individual score range for Depression, Anxiety, Stress scale is from 0 to 42 Maternal smoking is significantly associated with low family income (Chi-square P <.001) 94

109 Effect of Low Family Income on the Association Between Depression Scores and BMI To determine whether the influence of maternal smoking in pregnancy on the association between depression scores and BMI is a reflection of socio-economic status, we substituted low family income during pregnancy for maternal prenatal smoking. Table 4.5 provides a comparison of the characteristics according to maternal prenatal smoking and family income status. Compared to maternal prenatal non-smokers, maternal smokers were significantly younger (26.9 vs 29.5 yrs, P<0.001), had a lower family income (50.0 vs 30.6%, P<0.001), less tertiary education (34.5 vs 62.0%, P<0.001) and higher pregnancy life stress scores (1.6 vs 1.3, P<0.001). The offspring of mothers that smoked in pregnancy had a significantly higher BMI (26.1 vs 24.1 kg/m 2, P<0.001) and were more likely to be obese than offspring of mothers that did not smoke in pregnancy (20.4 vs 10.9%, P<0.001). Similar differences were observed comparing high versus low family income during pregnancy (Table 4.5). The multivariable regression model showed a similar pattern of results to those identified using maternal prenatal smoking. A significant interaction between low family income at pregnancy and depression-score (Interaction coefficient=0.091; 95%CI: 0.01, 0.17, P=0.027) indicated a positive association between BMI and depression score among offspring of mothers with a low family income during pregnancy (0.118; 95% CI 0.06, 0.18, P<.001), but no association in those with a high maternal family income during pregnancy (0.027; -0.03, 0.08, P=0.327)) (Table 4.6). This relationship was independent of gender and HC-use in females, maternal age, prepregnancy BMI, and maternal prenatal smoking. 95

110 Table 4.6. Multivariable adjusted BMI regression substituting family income at pregnancy for maternal prenatal smoking. Model Sample (N=1024) Slope Coefficient 95% CI P-value Depression-score ǂ , Low family income at pregnancy^ , Interaction Depression-score and Low income , Adjusted for the following variables HC-user female Ɨ , Male Ɨ , Maternal age (yr.) at pregnancy , Pre-pregnancy BMI (kg/m 2 ) , 0.53 <.001 Maternal smoking during pregnancy , 2.55 <.001 Model Constant , <.001 ǂ Depression score coefficient in offspring of high family income category Depression score coefficient in offspring of low family income category = ( ) = 0.118; 95%CI: 0.06, 0.18; P=<.001 (independent of gender, female HC-use, maternal age and pre-pregnancy BMI) ^Difference in BMI between offspring from low income vs high income families (at pregnancy) for those with depression-score = 0 Ɨ Female not using hormonal contraceptive is the reference 96

111 4.6 Discussion In the present study we found that the relationship between increasing depression scores and greater BMI of 20 year olds was modified by maternal prenatal smoking and low family income at pregnancy. The adverse effect of both maternal prenatal smoking and low income status on the depression-score and BMI association was independent of gender, maternal age, and maternal pre-pregnancy BMI. These results also took into account a variety of possible maternal and offspring confounders. We expected a positive association between depression-score and BMI given negative affect, a reflection of both anxiety and depression, and is known to be linked to obesity and both negative affect and obesity are on the rise. For example, the prevalence of negative mental affect in young people aged has risen to 30% among women and 14% among men in Sweden during [34]. A large proportion of young adults aged 20 and over in the United States are obese (34.6%) and 7.2% of US adults had depression, based on depressive symptoms experienced in the past 2 weeks [35]. Several studies have found cigarette smoking in pregnancy to be positively associated with offspring behavioural disorders [16,17,36,37]. However, the association of maternal prenatal smoking with offspring internalizing behaviours, such as depression, is not clear. In the Avon Longitudinal Study of Parents and Children the above association was not significant once adjusted for covariates such as socioeconomic status, parental psychopathology, and alcohol consumption [38]. Similarly, in the Generation R study, the association between maternal smoking in pregnancy and childhood internalizing behaviours was eliminated after adjustment for confounders such as parental educational level, family income, national origin, parental psychopathology, and child gender [15]. Maternal prenatal smoking has shared variance with other family psychosocial covariates of young offspring [39] and socioeconomic status may be one of them. Indeed, maternal smoking and maternal BMI have been suggested to be partly responsible for the intergenerational transmission of social inequalities in offspring BMI [40,41]. An in-utero effect of maternal prenatal smoking on offspring behavioural outcomes is plausible given the evidence that tobacco exposure during foetal life alters the epigenome [42], even in the absence of foetal growth restriction. Infants exposed to inutero tobacco smoke have significantly elevated adrenocorticotropin hormone levels 97

112 suggesting cigarette smoking may promote in-utero programming of the foetal hypothalamic pituitary-adrenal axis [43]. Similarly, in-utero tobacco exposure may predispose the offspring to childhood obesity through altered neurobehavioral impulse control systems and altered food satiation levels [44,45]. Apart from the above possible causal aspect of in-utero tobacco exposure, maternal prenatal smoking also has a strong psychosocial context. For example, continuing smoking in pregnancy is correlated with low maternal education [20], and children of women who smoke during pregnancy are more prone to social disadvantage [41]. Though there has been a decrease in the proportion of mothers smoking during pregnancy, the decline is more evident among highly educated women and less pronounced in younger and less educated women [46]. A low maternal education and a lower household income has been shown to contribute to the incidence of depression and development of obesity in adolescents [47]. The onset and persistence of depression among obese varies according to socioeconomic status as determined by income or educational level [6]. In this study, substituting family income at pregnancy for maternal smoking during pregnancy in the final model identified a significant association between depression score and BMI only for offspring of low income families. These findings suggest that the influence of maternal prenatal smoking is most likely a reflection of psychosocial influences and their effects on offspring predisposition for concurrent adiposity and depression. Limitations of the Study Our analyses took into account a variety of possible maternal and offspring confounders. A potential limitation of the study is that there was significant attrition from the original population cohort which included more high-risk pregnancies and low income families as they were recruited predominantly from public hospitals. However, the cohort now has a greater retention of socially advantaged families which would tend to underestimate the observed influences. In terms of generalizability of our findings, the proportion of mothers smoking in pregnancy was similar to various developed countries [15]. Considering that the DASS depression scores used in the study are dimensional, interpretability of clinical depression severity would need to be undertaken with caution [25]. However, some validity of the use of the depression scores in our 98

113 population was that in those who reported at some time having been diagnosed with depression by a medical practitioner, the DASS depression score was substantially higher than the remainder. In summary, we found a positive relationship between depression and BMI at age 20 that was independent of offspring gender, maternal age, pre-pregnancy BMI and other lifestyle covariates such as current smoking and alcohol drinking. Maternal prenatal smoking status modified this relationship, such that the association was evident only among offspring of mothers who smoked during pregnancy. Family income status at pregnancy, substituting for maternal prenatal smoking, also modified this relationship such that the association was also evident among offspring of mothers with a low family income at pregnancy. The results need to be seen in the context of a widening socioeconomic divide [23,48] and epidemic of worldwide levels of adiposity [1,49], with the likelihood of an accompanying increase in the levels of affective disorders such as depression. Identifying those most at risk of the co-association of adiposity and depressive symptoms is a public health priority which will need to be tackled at a societal level. 99

114 4.7 References 1. Kelly AS, Barlow SE, Rao G, et al. Severe obesity in children and adolescents: Identification, associated health risks, and treatment approaches: a scientific statement from the American Heart Association. Circulation. 2013;128(15): Olfson M, Blanco C, Wang S, Laje G, Correll CU. National trends in the mental health care of children, adolescents, and adults by office-based physicians. JAMA Psychiatry. 2014;71(1): de Wit L, Luppino F, van Straten A, Penninx B, Zitman F, Cuijpers P. Depression and obesity: a meta-analysis of community-based studies. Psychiatry Research. 2010;178(2): Gariepy G, Nitka D, Schmitz N. The association between obesity and anxiety disorders in the population: a systematic review and meta-analysis. International Journal of Obesity. 2010;34(3): Luppino FS, de Wit LM, Bouvy PF, et al. Overweight, obesity, and depression: a systematic review and meta-analysis of longitudinal studies. Archives of General Psychiatry. 2010;67(3): Williams EP, Mesidor M, Winters K, Dubbert PM, Wyatt SB. Overweight and obesity: prevalence, consequences, and causes of a growing public health problem. Current Obesity Reports. 2015;4(3): Hasler G, Drevets WC, Manji HK, Charney DS. Discovering endophenotypes for major depression. Neuropsychopharmacology. 2004;29(10): Dunkel Schetter C, Tanner L. Anxiety, Depression and stress in pregnancy: implications for mothers, children, research, and practice. Current Opinion in Psychiatry. 2012;25(2): Betts KS, Williams GM, Najman JM, Alati R. The relationship between maternal depressive, anxious, and stress symptoms during pregnancy and adult offspring behavioral and emotional problems. Depression and Anxiety. 2015;32(2): Korczak DJ, Lipman E, Morrison K, Duku E, Szatmari P. Child and adolescent psychopathology predicts increased adult body mass index: results from a prospective 100

115 community sample. Journal of Developmental & Behavioral Pediatrics. 2014;35(2): Louise S, Warrington NM, McCaskie PA, et al. Associations between anxiousdepressed symptoms and cardiovascular risk factors in a longitudinal childhood study. Preventive Medicine. 2012;54(5): Marmorstein NR, Iacono WG, Legrand L. Obesity and depression in adolescence and beyond: reciprocal risks. International Journal of Obesity. 2014;38(7): Oken E, Levitan EB, Gillman MW. Maternal smoking during pregnancy and child overweight: systematic review and meta-analysis. International Journal of Obesity. 2007;32(2): Al Mamun A, Lawlor DA, Alati R, O'Callaghan MJ, Williams GM, Najman JM. Does maternal smoking during pregnancy have a direct effect on future offspring obesity? Evidence from a prospective birth cohort study. American Journal of Epidemiology. 2006;164(4): Roza SJ, Verhulst FC, Jaddoe VW, et al. Maternal smoking during pregnancy and child behaviour problems: the Generation R study. International Journal of Epidemiology. 2009;38(3): Moylan S, Gustavson K, Øverland S, et al. The impact of maternal smoking during pregnancy on depressive and anxiety behaviors in children: the Norwegian mother and child cohort study. BMC Medicine. 2015;13: Ekblad M, Gissler M, Lehtonen L, Korkeila J. Prenatal smoking exposure and the risk of psychiatric morbidity into young adulthood. Archives of General Psychiatry. 2010;67(8): Miech RA, Caspi A, Moffitt TE, Wright BRE, Silva PA. Low socioeconomic status and mental disorders: a longitudinal study of selection and causation during young adulthood. American Journal of Sociology. 1999;104(4): Akee R, Simeonova E, Copeland W, Angold A, Costello EJ. Young adult obesity and household income: effects of unconditional cash transfers. American Economic Journal: Applied Economics. 2013;5(2):

116 20. Smedberg J, Lupattelli A, Mårdby A-C, Øverland S, Nordeng H. The relationship between maternal depression and smoking cessation during pregnancy: a cross-sectional study of pregnant women from 15 European countries. Archives of Women's Mental Health. 2015;18(1): Kipping RR, Smith M, Heron J, Hickman M, Campbell R. Multiple risk behaviour in adolescence and socio-economic status: findings from a UK birth cohort. The European Journal of Public Health. 2015;25(1): Hemmingsson E. A new model of the role of psychological and emotional distress in promoting obesity: conceptual review with implications for treatment and prevention. Obesity Reviews. 2014;15(9): Frederick CB, Snellman K, Putnam RD. Increasing socioeconomic disparities in adolescent obesity. Proceedings of the National Academy of Sciences. 2014;111(4): Newnham JP, Evans SF, Michael CA, Stanley FJ, Landau LI. Effects of frequent ultrasound during pregnancy: a randomised controlled trial. The Lancet. 1993;342(8876): Lovibond SH, Lovibond PF. Manual for the Depression Anxiety Stress Scales. (2nd Ed) Sydney: Psychology Foundation Tully PJ, Zajac IT, Venning AJ. The structure of anxiety and depression in a normative sample of younger and older Australian adolescents. Journal of Abnormal Child Psychology. 2009;37(5): Henry JD, Crawford JR. The short-form version of the Depression Anxiety Stress Scales (DASS-21): construct validity and normative data in a large non-clinical sample. British Journal of Clinical Psychology. 2005;44(2): Brunner Huber LR. Validity of self-reported height and weight in women of reproductive age. Maternal and Child Health Journal. 2007;11(2): Brennan-Jones CG, Eikelboom RH, Jacques A, et al. Protective benefit of predominant breastfeeding against otitis media may be limited to early childhood: results from a prospective birth cohort study. Clinical Otolaryngology. 2016:doi: /coa

117 30. Bhat SK BL, Robinson M, Burrows S, Mori T. Contrasting effects of prenatal life stress on blood pressure and BMI in young adults. Journal of Hypertension. 2015;33(4): Epstein NB, Baldwin LM, Bishop DS. The McMaster Family Assessment Device. Journal of Marital and Family Therapy. 1983;9(2): Jurakić D, Pedišić Ž, Andrijašević M. Physical activity of Croatian population: cross-sectional study using International Physical Activity Questionnaire. Croatian Medical Journal. 2009;50(2): Friedman MA, Brownell KD. Psychological correlates of obesity: moving to the next research generation. Psychological Bulletin. 1995;117(1): Lager A, Berlin M, Heimerson I, Danielsson M. Young people s health: health in Sweden: the National Public Health Report Chapter 3. Scandinavian Journal of Public Health. 2012;40(9 suppl): Pratt L, Brody D. Depression and obesity in the US adult household population, NCHS Data Brief. 2014(167): Cornelius MD, Goldschmidt L, Day NL. Prenatal cigarette smoking: long-term effects on young adult behavior problems and smoking behavior. Neurotoxicology and Teratology. 2012;34(6): Menezes AMB, Murray J, László M, et al. Happiness and depression in adolescence after maternal smoking during pregnancy: birth cohort study. PLoS ONE. 2013;8(11):e Brion M-J, Victora C, Matijasevich A, et al. Maternal smoking and child psychological problems: disentangling causal and non-causal effects. Pediatrics. 2010;126(1):e57-e Robinson M, McLean NJ, Oddy WH, et al. Smoking cessation in pregnancy and the risk of child behavioural problems: a longitudinal prospective cohort study. Journal of Epidemiology and Community Health. 2010;64(7): Koupil I, Toivanen P. Social and early-life determinants of overweight and obesity in 18-year-old Swedish men. International Journal of Obesity. 2007;32(1):

118 41. Fergusson DM, Woodward LJ, Horwood L. Maternal smoking during pregnancy and psychiatric adjustment in late adolescence. Archives of General Psychiatry. 1998;55(8): Ivorra C, Fraga MF, Bayón GF, et al. DNA methylation patterns in newborns exposed to tobacco in utero. Journal of Translational Medicine. 2015;13: McDonald SD, Walker M, Perkins SL, et al. The effect of tobacco exposure on the fetal hypothalamic-pituitary-adrenal axis. British Journal of Obstetrics and Gynaecology. 2006;113(11): Toschke A, Koletzko B, Slikker W, Hermann M, von Kries R. Childhood obesity is associated with maternal smoking in pregnancy. European Journal of Pediatrics. 2002;161(8): von Kries R, Toschke AM, Koletzko B, Slikker W. Maternal smoking during pregnancy and childhood obesity. American Journal of Epidemiology. 2002;156(10): Rogers JM. Tobacco and pregnancy: Overview of exposures and effects. Birth defects research part C: Embryo Today: Reviews. 2008;84(1): Yackobovitch-Gavan M, Meshy-Tamir R, Nagelberg N, Phillip M, Meyerovitch J. Psychosocial factors associated with depressive mood in Israeli obese adolescents. Journal of Health Psychology. 2014;19(4): Chung A, Backholer K, Wong E, Palermo C, Keating C, Peeters A. Trends in child and adolescent obesity prevalence in economically advanced countries according to socioeconomic position: a systematic review. Obesity Reviews. 2016;17(3): Ogden CL, Carroll MD, Lawman HG, et al. Trends in obesity prevalence among children and adolescents in the united states, through JAMA. 2016;315(21):

119 Chapter 5 Relationships between Depression and Anxiety Symptoms Scores and BP in Young Adults 5.1 Preamble This chapter examined blood pressure levels of the Raine Study participants in context of prevalence of depression/anxiety symptoms at age 20. The study examined the BP and cardiovascular risk factor profiles of the cohort, and demonstrated that offspring phenotype and behavioural factors influence the association between BP and depression or anxiety scores. This chapter additionally demonstrated the interaction effect of higher BMI levels on the association between offspring self-reported depression and BP at age 20. By showing the varying influence of higher BMI levels compared to lower BMI levels on this association, it addressed a theme central to this thesis that the young adult phenotypic and behavioural differences could be important determinants for CV risk factors such as depression and hypertension. 105

120 5.2 Abstract Depression and anxiety are risk factors for cardiovascular disease but their relationship to blood pressure (BP) is less clear. Age related comorbidity and lifestyle factors may confound these relationships. This study aimed to assess the relationships between BP, depression and anxiety symptom scores and self-reported history of depression in young adults. Data on 1014 participants aged 20 years from the Western Australian Cohort (Raine) Study were analyzed for cross-sectional associations between clinic BP and Depression, Anxiety, Stress Scale (DASS) questionnaire scores or a reported history of depression, taking into account relevant confounders. Multivariable adjusted analyses showed an inverse relationship between systolic BP, but not diastolic BP, with depression (coefficient=-0.10; P=0.012) and anxiety (coefficient=-0.13; P=0.018) scores, independent of gender, BMI, female hormonal contraceptive use, alcohol consumption, birth weight and maternal hypertension in pregnancy. Systolic BP was 1.6 mmhg lower for 2SD (16 units) increase in depression score. There was an inverse association between self-reported history of depression (15.8% of participants) and systolic BP (coefficient=-1.91; P=0.023) with an interaction with increasing BMI (Interaction coefficient=-0.43; P=0.002) enhancing this difference. Our findings show that systolic BP in young adults is inversely associated with depression or anxiety scores, independent of a range of lifestyle confounders. Despite a positive association between BMI and BP, adiposity enhanced the inverse association between self-reported history of depression and systolic BP. These findings contrast with the predisposition of depressed subjects to cardiovascular disease in later life when decades of unhealthy lifestyle changes may dominate. Keywords: depression; blood pressure; obesity; Raine-Study 106

121 5.3 Introduction Hypertensive individuals tend to be hyper-responsive to acute stressors in terms of both BP and increases in total peripheral resistance, suggesting that BP is responsive to the psychosocial environment. Depression has a high prevalence and can adversely affect the course of hypertension [1]. Furthermore, clinical depression is a significant risk factor for mortality in patients with coronary heart disease [2]. Similarly, a history of a lifetime combination of anxiety and depression has been associated with increased risk of cardiovascular disease [3]. Large population databases indicate significant psychiatric comorbidities in patients with hypertension [4]. However, studies examining the relationship between depression or anxiety symptoms with BP have variably reported positive [5-10], negative [11-16] or non- existent [17-20] associations. Meng et al7 reported a meta-analysis showing that depression increased the risk of hypertension incidence and suggested that it is an independent risk factor of hypertension. One of the few reports using ambulatory BP found a positive association between depressive symptoms scores and daytime BP [21]. Mild depression in a Japanese population was also associated with higher systolic BP based on 7-day/24-hr ambulatory BP monitoring [22] and hypertension in elderly adults was related to symptoms of significant depression but not to generalized anxiety [9]. In contrast, several studies using a variety of questionnaire based scores of depression or anxiety symptoms have shown inverse associations with BP, including the large population-based HUNT study in Norway [13-15] and the 1946 British Medical Research Council National Survey of Health and Development [11]. Low BP has also been reported with depression or depressive symptoms among the elderly [23,24]. A number of studies have failed to show any relationship between depression or anxiety and BP or risk of hypertension [17,19,20]. Differences between studies could be due to differences in the populations selected, age related factors, anti-depressant or antihypertensive treatment, and varying means of assessing depression or anxiety. Other important sources of heterogeneity include different outcome measures such as systolic and diastolic BP, BP changes or hypertension. We previously reported an inverse association between anxious and depressed scores and systolic BP limited to 14-year-old boys in the Western Australian Pregnancy Cohort (Raine) Study [12]. The cohort is one of the longest running prospective studies on child 107

122 health and development that offers a wide range of lifestyle factors relating to mothers and their offspring [25]. The present study examined the association between depression or anxiety symptoms and BP in 20-year-old Raine study participants, an age when they have established adult behaviours, but with little or no co-morbidity. As adiposity is associated with depression [26] and BP [27] we also examined for interactions between BP and body mass, depression or anxiety symptoms or a reported history of depression. 108

123 5.4 Material and methods The Raine Study is a prospective longitudinal pregnancy study of 2868 live births from pregnant women enrolled at 18 weeks of gestation [25]. Briefly, 2900 pregnant women were recruited between weeks gestation through the public antenatal clinic at King Edward Memorial Hospital in Perth, Western Australia, or surrounding private practices between The study has approval from the Human Research Ethics Committees at King Edward Memorial Hospital, Princess Margaret Hospital for Children and The University of Western Australia. This analysis uses data from the 20- year follow-up of the offspring, conducted from Informed written consent was obtained from the parents or a primary care giver at each follow- up and from the participants at 20 years. BP and Anthropometry Resting supine BP was recorded using an oscillometric sphygmomanometer (Dinamap ProCare 100; Soma Technology, Bloomfield, Connecticut), with an appropriate cuff size. After resting quietly in a supine position for 5 minutes, six BP recordings were taken sequentially every 2 minutes and the last five were averaged. Systolic prehypertension (SBP-PH) and hypertension (SBP-H) were defined using the adult Joint National Committee on the Prevention, Detection, Evaluation, and Treatment of High BP (JNC-7) criteria ( SBP-PH as mmhg and SBP-H as 140 mmhg. Weight was recorded with a Wedderburn Chair Scale (nearest 100g) after fasting and wearing minimal clothing. Height was measured without shoes using a Holtain Stadiometer, Holtain Ltd, Crosswell, UK (nearest 0.1cm). DASS Depression-score The Depression, Anxiety, Stress Scale (DASS-21) is the short-form of the DASS-42, a combination of three self-report scales and comprising twenty-one questions. Seven questions in each category measure the negative emotional states respectively of depression, anxiety and stress [28]. The DASS-21 scores were multiplied by 2 as per scoring instructions allowing comparison with DASS-42 normative data [28]. DASS-21 correlates highly with other validated measures of depression and anxiety [28,29]. Participants were also asked if they had ever been diagnosed with depression asking Do you have now, or have you had in the past, any of the following health professional 109

124 diagnosed medical conditions or health problems?. We used continuous depression or anxiety scores or self-reported clinical depression, each analyzed separately. Lovibond et al [28] have shown individuals with depression scores 9 have no clinical depression and scores >9 associate with depression. Thus, baseline characteristics of participants are also presented based on these criteria. Maternal covariates Maternal hypertension comprised hypertensive disorders during pregnancy including gestational hypertension, pre-eclampsia, pre-pregnancy hypertension, and prepregnancy hypertension with superimposed pre-eclampsia. The pre-pregnancy BMI was calculated from self-reported weight and height measured at 18-weeks gestation. If women could not answer this question, their weight at the completion of this questionnaire was used to estimate pre-pregnancy weight. Self- report of weight by women of reproductive age has been found to be valid with 84% of women correctly classified into the correct BMI categories [30]. Maternal self-reported data for weekly alcohol intake were obtained at 18 weeks' gestation. Maternal prenatal smoking was self-reported for daily smoking. Maternal smoking and alcohol consumption in pregnancy were dichotomized to yes/no responses. Mothers with higher education were ranked as having a tertiary education. Annual family income at the time of pregnancy ( ) was dichotomized, with low income status defined as <$24,000 Australian- Dollars (AUD), in accordance with the poverty line at the time. Breastfeeding data from questionnaires completed by mothers assessed the age breastfeeding stopped. At 18-weeks gestation mothers were asked whether they had experienced any of ten life-stress events since becoming pregnant [31]. Familyfunctioning at adolescent age 14yr was determined using the General Functioning Scale from the McMaster Family Assessment Device [32]. Offspring covariates Birth weight of the participants was obtained from hospital records. Self-reported measures of smoking and alcohol consumption at 20 years were dichotomized to a yes/no response. Female hormonal contraceptive (HC) use was based on self-reported current use of the oral contraceptive (OC) pill, implant, injection or any intrauterine HC-device. The short form of the International Physical Activity Questionnaire was 110

125 used to derive a measure of metabolic equivalents (METS), a unit of resting metabolic rate [33]. Statistical analyses Analyses were performed using STATA 12 (StataCorp, Texas, USA). Continuous variables are presented as mean ± standard deviation (SD). Categorical variables are shown as frequencies (percentages). Categorical variables were compared using the χ2 test and t-test was used to examine significant difference in characteristics across groups. Ordinary least square models regressed the BP outcome on each of DASS- Depression, DASS-anxiety scores and self-reported clinical depression. The linearity of the relation with the DASS scores was assessed examining scatterplot overlaid with locally weighted regression smoother. Depression scores were regressed on selfreported depression to determine if current scores were related to prior diagnosis. Preliminary univariate analyses involved variables of clinical and biological relevance for systolic and diastolic BP [34]. The offspring covariates included gender, gestational age at delivery (weeks), birth weight, breast feeding ( 4 months), current alcohol drinking, current smoking and physical activity (hours/week). Maternal covariates included age, history of hypertension in pregnancy, pre-pregnancy BMI, smoking, alcohol consumption and life-stress score during pregnancy, post-school higher (tertiary) education and low family income at the time of pregnancy. Multivariable analyses were conducted separately for depression or anxiety scores or self- reported clinical depression. Variables potentially associated (P<0.10) with BP in univariate analyses were included in the initial multivariable regression model. A backward elimination regression identified the final set of significant (P<0.05) covariates. Covariates with the largest P-value above 0.05 were excluded first and at each step the model was examined for evidence of the excluded covariate s importance from the change in regression coefficient and model variance. Given few studies in the literature show gender differences in BP and depression scores, we analyzed for possible gender interactions for the relation between depression score or self-reported depression and BP. Obesity has been found to be associated with both BP and depression, so the interaction of depression and BMI was also investigated. Meancentering was applied to BMI where the overall mean of BMI was subtracted from individual BMI of the participants. This shifts the distribution so that mean is zero and allows meaningful interpretation of regression coefficients in models where interaction 111

126 with BMI is present. All regression models used within family cluster adjustment to correct for correlation between a small number of siblings. 5.5 Results At 20 years of age 1348 participants attended the follow-up (Figure 5.1). Data on offspring primary variables of interest including BP, depression/anxiety scores, BMI and gender were available for 1078 participants. Taking into account all confounders, data on 1014 participants was available for analysis. There were 21 offspring with siblings included. Figure 5.1. Flow diagram of Raine Study participants attending the 20-year follow-up. A comparison of the participants in the final model (n=1014) with those that did not participate or had incomplete data in the 20-year follow-up (n=1854) shows significantly more mothers of those that did not participate smoked during the first trimester of pregnancy (31.4 vs 19.1%), were less likely to have a tertiary education (43.3 vs 57.2%) or to consume alcohol during pregnancy (42.8 vs 50.3%), were younger (26.9 vs 29.0 years) and more likely to have a low family income at the time of pregnancy (49.7 vs 33.9%) (Table 5.1). 112

127 Table 5.1. Pregnancy related characteristics for mothers of participants compared with mothers of non-participants in the 20-year follow-up Mothers of Mothers of Characteristics participants that did not participate participants included P value (n = 1854) (n=1014) Gestational age (weeks) 38.5 ± ± Offspring birth weight (kg) 3.3 ± ± Age (yr.)* 26.9 ± ± 5.6 <0.001 Pre-pregnancy BMI (kg/m 2 )* 22.4 ± ± Pre-pregnancy Obesity, n (%)* 125 (7.0) 59 (5.8) Hypertension in pregnancy, n (%)* 466 (26.0) 254 (25.0) Smoking in pregnancy, n (%)* 562 (31.4) 194 (19.1) <.001 Tertiary education, n (%)* 776 (43.3) 580 (57.2) <.001 Low family income at pregnancy, n (%)* 821 (49.7) 334 (33.9) <.001 Alcohol drinker in pregnancy, n (%)* 766 (42.8) 510 (50.3) <.001 Values are arithmetic means ± SD or number of individuals (%) * Data available for 2804 mothers These findings are indicative of some retention bias for families more likely to be welleducated, health conscious and having a relatively higher income. Young adult offspring that did not complete the DASS Depression score questionnaire (n=242) were more likely to have mothers with a low family income at the time of pregnancy (43.2 vs 34.4%; P=0.012) and more likely to be males (70.2 vs 47.8%; P<0.001) (not shown). 113

128 Maternal and Offspring Cohort Characteristics Male offspring had a significantly higher systolic BP and a greater percentage were pre- hypertensive or hypertensive compared with females (Table 5.2). Table 5.2. Participants characteristics Offspring characteristics at 20Yr Total (n=1078) Females (n=563) Males (n=515) P value Systolic BP (mmhg) ± ± ± 11.8 <0.001 Diastolic BP (mmhg) 65.4 ± ± ± Hypertension 140mmHgǂ, n (%) 41 (3.8) 3 (0.5) 38 (7.4) <0.001 Prehypertension mmHgǂ, n (%) 359 (33.0) 99 (17.6) 260 (50.5) <0.001 Heart rate (bpm) 77.3 ± ± ± 11.3 <0.001 BMI (kg/m 2 ) 24.4 ± ± ± BMI <25, n (%) 723 (67.1) 384 (68.2) 339 (65.8) BMI 25 and <30, n (%) 219 (20.3) 99 (17.6) 120 (23.3) BMI 30, n (%) 136 (12.6) 80 (14.2) 56 (10.9) Depression score Ɨ 7.0 ± ± ± 7.3 <0.001 Anxiety score Ɨ 5.0 ± ± ± 4.6 <0.001 Stress score Ɨ 8.9 ± ± ± 7.2 <0.001 Total DASS score 21.0 ± ± ± 16.8 <0.001 Self-reported depression, n (%) 166 (15.8) 109 (19.9) 57 (11.4) <0.001 Hormonal contraceptive use, n (%) (60.4) - - Alcohol drinker, n (%) 721 (69.3) 358 (65.9) 363 (73.0) Current Smokers, n (%) 156 (14.5) 74 (13.1) 82 (15.9) Physical activity (METS-hrs./week) 59.4 ± ± ± 75.5 <0.001 Birth weight (kg) 3.3 ± ± ± 0.6 <0.001 Breast fed 4 months, n (%) 701 (68.5) 365 (67.8) 336 (69.1)

129 Maternal characteristics # Age(yr.) 29.0 ± ± ± Pre-pregnancy BMI (kg/m 2 ) 22.2 ± ± ± Pre-pregnancy Obesity, n (%) 63 (6.0) 31 (5.6) 32 (6.4) Hypertension in pregnancy, n (%) 263 (25.0) 123 (22.3) 140 (27.9) Smoking in pregnancy, n (%) 206 (19.6) 122 (22.1) 84 (16.8) Tertiary education, n (%) 596 (56.6) 306 (55.4) 290 (57.9) Pregnancy life stress score 1.4 ± ± ± Low income in pregnancy, n (%) 351 (34.4) 186 (35.0) 165 (33.7) Family functioning score 26.9 ± ± ± Values are arithmetic means ± SD or number of individuals (%) ǂ Hypertension and prehypertension according to JNC-7 criteria METS, Metabolic equivalents ƗScore range for each Depression, Anxiety, Stress scale 0-42 # Complete data for maternal characteristics were available for 1053 participants <$24,000 (AUD) during Diastolic BP was not different between sexes but heart rate was lower in males (Table 5.2). Although BMI was not different between the genders, males were more likely to be categorized as overweight. The average depression, anxiety and stress scores were lower in males, and the proportion with self-reported depression was significantly higher among females compared to males (19.9 vs. 11.4%). Males had a significantly higher birth weight, and were more likely to be physically active, and consume alcohol. The proportion of HC-use by females was 60.4% and 15.8% of all participants reported having at some time had a diagnosis of depression (Table 5.2). In these subjects the DASS-Depression scores were 8.34 units (95%CI: 6.61, 10.10, P<0.001) higher than those without a history of depression (not shown). Participants with depression (score >9) compared to those with no depression (score 9) had a significantly higher BMI (25.0 vs.24.2 kg/m2) and were more likely to be obese (17.5 vs. 10.7%). They had higher anxiety (9.5 vs. 3.2) and stress (16.1 vs.6.0) scores, they were more likely to be female (59.2 vs. 49.4%) and current smokers (19.7 vs. 12.4), but less physically active (49.7 vs METS-hrs./week) and with a lower family functioning score (26.0 vs.27.2) (Table 5.3). 115

130 Table 5.3. Participant characteristics according to presence or absence of depression No Depression Depression P Offspring characteristics at 20Yr (Depression-score 9) (Depression-score >9) value (n=769) (n=309) Gender - Female, n (%) 380 (49.4) 183 (59.2) Hormonal contraceptive use among females, n (%) 228 (60.0) 112 (61.2) Systolic BP (mmhg) ± ± Diastolic BP (mmhg) 65.4 ± ± Hypertension 140mmHgǂ, n (%) 31 (4.0) 10 (3.2) Prehypertension mmHgǂ, n (%) 267 (34.7) 92 (29.8) Heart rate (bpm) 76.9 ± ± BMI (kg/m 2 ) 24.2 ± ± BMI <25, n (%) 529 (68.8) 194 (62.8) BMI 25 and <30, n (%) 158 (20.5) 61 (19.7) BMI 30, n (%) 82 (10.7) 54 (17.5) Anxiety score Ɨ 3.2 ± ± 7.2 <0.001 Stress score Ɨ 6.0 ± ± 8.3 <0.001 Self-reported depression, n (%) 65 (8.8) 101 (33.1) <0.001 Alcohol drinker, n (%) 517 (70.1) 204 (67.3) Current Smokers, n (%) 95 (12.4) 61 (19.7) Physical activity (METS-hrs./week) 63.3 ± ± Birth weight (kg) 3.3 ± ± Breast fed 4 months, n (%) 502 (68.8) 199 (67.7) Maternal characteristics # 116

131 Age(yr.) 28.8 ± ± Pre-pregnancy BMI (kg/m 2 ) 22.2 ± ± Pre-pregnancy Obesity, n (%) 40 (5.3) 23 (7.6) Hypertension in pregnancy, n (%) 199 (26.5) 64 (21.2) Smoking in pregnancy, n (%) 142 (18.9) 64 (21.1) Tertiary education, n (%) 425 (56.7) 171 (56.4) Pregnancy life stress score 1.4 ± ± Low income in pregnancy, n (%) 239 (32.7) 112 (38.6) Family functioning score 27.2 ± ± Values are arithmetic means ± SD or number of individuals (%) ǂ Hypertension and prehypertension according to JNC-7 criteria METS, Metabolic equivalents Ɨ Score range for Anxiety and Stress scale 0-42 # Complete data for maternal characteristics were available for 1053 participants <$24,000 (AUD) during Univariate regression DASS-depression and DASS-anxiety scores were inversely associated with offspring systolic BP (coefficient=-0.13; P=0.005 and coefficient=-0.20; P=0.008, respectively) but not diastolic BP (Table 5.4). Offspring systolic BP associated positively with male gender, current alcohol consumption and physical activity, as well as maternal pre-pregnancy BMI and obesity (Table 5.4). Offspring BMI and obesity, and maternal hypertension in pregnancy were positively associated with both systolic and diastolic BP at age 20 (Table 5.4). 117

132 Table 5.4. Univariate BP regression. Explanatory variables Systolic BP Diastolic BP Offspring variables Coef. 95%CI P-value Coef. 95%CI P-value Depression-score , , Anxiety-score , , Gender (Female reference) Male , < , level Gender (No-HC-user reference) Female HC-user , , Male , < , BMI at 20yr , 0.84 < , Obese at 20yr , 9.18 < , Birth weight (kg) , , Breast feeding 4months , , Alcohol drinking at 20yr , , Smoking at 20yr , , Physical activity (METShrs./week) Maternal variables , 0.04 < , Maternal age(yr.) , , Hypertension , 5.69 < , Pre-pregnancy BMI (kg/m 2 ) , , Pre-pregnancy obese , , Alcohol intake , , Maternal smoking , , Life stress score , , Low family income , , Maternal tertiary education , ,

133 Multivariable analyses The inverse association between systolic BP and depression-score remained significant (P=0.012), independent of gender and BMI, female HC-use, alcohol consumption, birth weight and maternal hypertension in pregnancy (Table 5.5). A 2SD (16 units) increase in depression-score was associated with a 1.6 mmhg reduction in systolic BP. There were no significant interactions between depression-score and gender (Interaction coefficient=0.01; P=0.921) or depression-score and BMI (Interaction coefficient=-0.01; P=0.392) (not shown). Table 5.5. Multivariable adjusted systolic BP regression on depression-score (Final Model) N=1014; Model R 2 =0.327* Coefficient 95%CI P-value Depression-score , Adjusted for the following variables Female HC-user Ɨ , Male Ɨ , <.001 BMI at 20yr (kg/m 2 ) , 0.90 <.001 Alcohol drinking at 20yr , Birth weight (kg) , Maternal hypertension in pregnancy , Model Constant , <.001 ƗReference: Female not using hormonal contraception * The model explains 32.7% of the variance in SBP There was no significant association between anxiety-score and systolic BP (coefficient=-0.08; P=0.176) in multivariable regression analysis that adjusted for gender and BMI, female HC use, alcohol consumption, birth weight and maternal hypertension in pregnancy (Table 5.6). 119

134 A scatter plot of anxiety-score and systolic BP identified two outliers with systolic BP of 158 and 157 mmhg. Removing these outlying observations resulted in a significant inverse association between anxiety-score and systolic BP (coefficient=-0.13; P=0.018) (Table 5.6). Table 5.6. Multivariable adjusted systolic BP regression on anxiety-score Full sample (n=1014)* Final Model (n=1012)** Coef. 95%CI P- value Coef. 95%CI P- value Anxiety-score , , Adjusted for the variables Female HC-user Ɨ , , Male Ɨ , , <.001 BMI at 20yr (kg/m 2 ) , 0.88 < , 0.87 <.001 Alcohol drinking at 20yr , , Birth weight (kg) , , Maternal hypertension in pregnancy , , Model Constant , < , <.001 ƗReference: Female not using hormonal contraception * The model explains 32.4% of the variance in SBP (R 2 =0.324) ** The model, without two outliers with SBP of 158 mmhg and 157 mmhg, explains 32.4% of the variance in SBP (R 2 =0.324) There was a significant inverse association between self-reported history of depression and systolic BP (coefficient=-1.91; P=0.023) (Table 5.7), with a significant interaction between self-reported depression and BMI (coefficient=-0.43; P=0.002). The interaction indicated that the higher the BMI, the more inverse the relationship between the selfreported depression and SBP. 120

135 The association between self-reported depression and systolic BP was independent of female HC-use, offspring alcohol drinking, birth weight and maternal hypertension in pregnancy (Table 5.7). Table 5.7. Multivariable adjusted systolic BP regression on self-reported clinical depression Model R 2 =0.328* Coef. 95%CI P-value Self-Reported history of depression ǂ , Mean-centered BMI at 20yr (kg/m 2 )^ , 1.00 <.001 Interaction Self-Reported depression and mean-centered BMI , Adjusted for the following variables Female HC-user Ɨ , Male Ɨ , <.001 Alcohol drinking at 20yr , Birth weight (kg) , Maternal hypertension in pregnancy , 4.01 <.001 Model Constant , <.001 ǂ Difference in SBP between offspring having self-reported clinical depression vs nondepression for an average BMI ^BMI coefficient in offspring having no depression BMI coefficient in offspring having self-reported clinical depression = ( ) = 0.414; 95% CI ; P<0.001 (independent of gender, female HC-use, offspring alcohol drinking, birth weight and maternal hypertension in pregnancy) Ɨ Reference: Female not using hormonal contraceptives *The model explains 32.8% of the variance in SBP 121

136 None of the participants used antidepressants at age 20. However, nine participants were taking medication that could affect BP. Although the small number of participants taking medication would not provide an accurate estimate of the influence of medication, a sensitivity analysis showed that removing these 9 subjects had no influence on the parameter estimates (not shown). There was no significant interaction between gender and self-reported history of depression (Table 5.8). Table 5.8. Systolic BP regression on self-reported history of depression and gender interaction (N=1022; R 2 =0.311) * Coefficient 95%CI P-value Self-Reported history of depression ǂ , Male Ɨ , <.001 Self-Reported depression and gender interaction # , Mean-centered BMI at 20yr (kg/m 2 ) , 0.86 <.001 Model Constant , <.001 * The model explains 31.1% of the variance in SBP ǂ Coefficient of self-reported Depression and SBP in females. # Equates to regression coefficient= (95%CI: -6.07, 0.43) for the association between self-reported depression and SBP in males. Ɨ Baseline reference group: Female 122

137 5.6 Discussion Analysis of data on young adults from the Raine Study has shown an inverse association between depression or anxiety scores and systolic BP independent of possible confounders. The magnitude of the BP reduction was similar to that of the BP elevation associated with alcohol consumption and HC-use in women. Participants with a history of self-reported depression also showed an inverse association with BP, independent of lifestyle confounders. However, there was an interaction with adiposity such that the inverse association between self-reported depression and systolic BP was accentuated with increasing BMI. Participants with a history of self-reported depression also had substantially higher depression symptom scores. The inverse association between depression or anxiety scores and systolic BP agrees with some of the cross-sectional and prospective studies in older populations [11,16]. The Norwegian Hunt study reported an inverse association between depression or anxiety scores [13-15] in both men and women approximately one decade older than our cohort. They were also able to exclude effects of antihypertensive or antidepressant therapy. Confounding effects of antidepressant therapy in an older population with clinically diagnosed depression have been suggested from the Netherlands Study of Depression and Anxiety which showed that remitted and currently depressed participants, even after correcting for antidepressant use, had significantly lower systolic BP and were less likely to have isolated systolic hypertension than controls [16], whereas tricyclic antidepressant use, selective serotonin reuptake inhibitors, and noradrenergic and serotonergic working antidepressants associated with hypertension stage-1 [16]. The present study in 20-year-olds showed approximately 50% of males were classified as pre-hypertensive and they were more likely to be hypertensive than females. These findings are not unexpected given the arbitrary definition for pre-hypertension (systolic BP mmhg) is what used to be considered normal BP and the fact that males have substantially higher systolic BP than females from early childhood [35]. The multivariable model estimates using continuous BP showed no significant sex differences in the associations with depression or anxiety symptoms, despite the higher scores for depression in women and higher blood pressures in men. This is in agreement with studies in older adults [11,13-15], but in contrast with our previous finding in the Raine Cohort at 14 years of age where the relationship was only seen in boys [12]. 123

138 However in that report no gender and depression score interaction was undertaken at age 14 [12]. The Baltimore Longitudinal Study of Ageing [36] showed lower BP in men with higher depression scores, but the opposite relationship in younger women. Differing questionnaires to assess depression/anxiety may also account for some of these variations. Studies in older subjects reporting a positive association with depression scores or clinical depression and BP or suggesting recurrent depressive symptoms as a risk factor for hypertension [37], may have been confounded by effects of anti-depressant therapy [5,17], concomitant illness, and poor lifestyle predisposing to hypertension such as smoking, alcohol drinking, poor diet and physical inactivity [10,20]. However, in the young adults in our study the relation between depression scores and BP was independent of a range of lifestyle confounders. None of the participants in our study was taking antidepressant medication. We did not adjust for medications that could affect BP because a sensitivity analysis indicated that removing the nine participants taking these medications had no influence on the association between depression scores or self-reported depression and BP. Although unidentified confounders may explain our findings, we now have a pattern reported from adolescence [12] and 20-year-olds in our cohort, through to later adult life by others [11,16]. These inverse associations are surprising and counter-intuitive mechanistically, given that adiposity is a risk factor for both depression [38] and high BP [39]. Moreover, depression in young adults associates with increased markers of inflammation, hypothalamic-pituitary-adrenal axis dysfunction with hypercortisolism, decreased heart rate variability, elevated catecholamines and endothelial dysfunction [40], all of which would be expected to elevate BP [41,42]. These mechanisms do not accord with the negative associations between self-reported depression, depression or anxiety symptom scores and BP [13-15]. However, such mechanisms are plausible contributors to results of studies reporting an association between clinically significant depression symptoms and established hypertension in older adults in whom co-morbidities and unhealthy lifestyles associated with both conditions are likely to be well established [9]. We have no ready explanation for a causal relationship between depressive or anxiety tendencies and lower BP, however the findings, now consistently reported from childhood through to late adult life, suggest contrasting mechanisms underlying 124

139 depression and blood pressure control operating through any one of the central nervous system pathways involved in autonomic function, the hypothalamic/adrenal axis or leptin metabolism. Indeed, studies of catecholamine metabolism in patients with severe depression showed a bimodal pattern with an increased spill-over only in those with concomitant panic disorder [43]. Compared with healthy controls, depressed and anxious participants showed differential autonomic nervous system reactivity; hyporeactivity for a cognitively challenging stressor contrasting with hyper-reactivity for a personal-emotional stressor [44]. Thus, it is possible that sympathetic nervous system hypo-reactivity and/or increased vagal tone in depressed prone young adults may result in lower blood pressures. Another physiological pathway that may contribute to the relationship between depression and adiposity is leptin, a peptide hormone that acts not only as an anti-obesity hormone, but can influence emotion control and cognition [45]. Strengths of the study include use of continuous mood scores which avoids arbitrary cut offs of continuous mood traits, and accounting for a wide range of potential lifestyle confounders. When we did select those who reported a history of depression, the results were in accord with those using the depression-score. However, we do not know the basis on which the diagnosis of depression had been made. Limitations of the study include its observational and cross-sectional nature which cannot prove cause and effect. Nevertheless, the results concur with our earlier findings in the Raine cohort boys at 14 [12] and with prospective studies [11,13-16] suggesting a cumulative influence of depression or anxiety symptoms on lower BP. There was significant attrition from the original cohort which included more high-risk pregnancies and low income families as they were recruited predominantly from public hospitals. Therefore, the findings may not be generalizable to the general population of 20-year-olds. However, given that the cohort now has a greater retention of socially advantaged families who are less likely to be depressed and more likely to have lower BP; this would only tend to underestimate the strength of the observed associations at a population level. In conclusion, the present study has shown an inverse association between depression or anxiety symptom scores and systolic BP in a contemporary population at age 20, independent of a range of likely confounders. The influence of BMI in enhancing the inverse association between self-reported history of depression and systolic BP may have clinical importance and is of further interest. These findings contrast with the predisposition of depressed subjects to cardiovascular disease in later life when decades 125

140 of unhealthy lifestyle changes may dominate. It will be valuable to follow the evolution of these effects as the cohort ages. 5.7 References [1]. Scalco AZ, Scalco MZ, Azul JBS, Lotufo Neto F. Hypertension and depression. Clinics. 2005;60: [2]. Barth J, Schumacher M, Herrmann-Lingen C. Depression as a risk factor for mortality in patients with coronary heart disease: a meta-analysis. Psychosomatic Medicine. 2004;66(6): [3]. Strine TW, Mokdad AH, Balluz LS, et al. Depression and anxiety in the United States: findings from the 2006 behavioral risk factor surveillance system. Psychiatric Services. 2008;59(12): [4]. Sandström YK, Ljunggren G, Wändell P, Wahlström L, Carlsson AC. Psychiatric comorbidities in patients with hypertension-a study of registered diagnoses in the total population in Stockholm County, Sweden. Journal of Hypertension. 2016;34(3): ; discussion 420. [5]. Jackson CA, Pathirana T, Gardiner PA. Depression, anxiety and risk of hypertension in mid-aged women: a prospective longitudinal study. Journal of Hypertension. 2016;34(10): [6]. Cuffee Y, Ogedegbe C, Williams NJ, Ogedegbe G, Schoenthaler A. Psychosocial risk factors for hypertension: an update of the literature. Current Science Inc. 2014;16(10):1-11. [7]. Meng L, Chen D, Yang Y, Zheng Y, Hui R. Depression increases the risk of hypertension incidence: a meta-analysis of prospective cohort studies. Journal of Hypertension. 2012;30(5): [8]. Davidson K, Jonas BS, Dixon KE, Markovitz JH. Do depression symptoms predict early hypertension incidence in young adults in the cardia study? Archives of Internal Medicine. 2000;160(10):

141 [9]. Maatouk I, Herzog W, Böhlen F, et al. Association of hypertension with depression and generalized anxiety symptoms in a large population-based sample of older adults. Journal of Hypertension. 2016;34(9): [10]. Jani BD, Cavanagh J, Barry SJE, Der G, Sattar N, Mair FS. Relationship between blood pressure values, depressive symptoms, and cardiovascular outcomes in patients with cardiometabolic disease. The Journal of Clinical Hypertension. 2016;18(10): [11]. Tikhonoff V, Hardy R, Deanfield J, et al. Symptoms of anxiety and depression across adulthood and blood pressure in late middle age: the 1946 British birth cohort. Journal of Hypertension. 2014;32(8): [12]. Louise S, Warrington NM, McCaskie PA, et al. Associations between anxiousdepressed symptoms and cardiovascular risk factors in a longitudinal childhood study. Preventive Medicine. 2012;54(5): [13]. Hildrum B, Mykletun A, Stordal E, Bjelland I, Dahl AA, Holmen J. Association of low blood pressure with anxiety and depression: the Nord-Trøndelag Health Study. Journal of Epidemiology and Community Health (1979-). 2007;61(1): [14]. Hildrum B, Mykletun A, Holmen J, Dahl AA. Effect of anxiety and depression on blood pressure: 11-year longitudinal population study. The British Journal of Psychiatry. 2008;193(2): [15]. Hildrum B, Romild U, Holmen J. Anxiety and depression lowers blood pressure: 22-year follow-up of the population based HUNT study, Norway. BMC Public Health. 2011;11(1):601. [16]. Licht CMM, de Geus EJC, Seldenrijk A, et al. Depression is associated with decreased blood pressure, but antidepressant use increases the risk for hypertension. Hypertension. 2009;53(4): [17]. Delaney JAC, Oddson BE, Kramer H, Shea S, Psaty BM, McClelland RL. Baseline depressive symptoms are not associated with clinically important levels of incident hypertension during two years of follow-up: the multi-ethnic study of atherosclerosis. Hypertension. 2010;55(2):

142 [18]. Grimsrud A, Stein DJ, Seedat S, Williams D, Myer L. The association between hypertension and depression and anxiety disorders: results from a nationallyrepresentative sample of South African adults. PLoS One. 2009;4(5):e5552. [19]. Yan LL, Liu K, Matthews KA, Daviglus ML, Ferguson T, Kiefe CI. Psychosocial factors and risk of hypertension: the coronary artery risk development in young adults (cardia) study. JAMA. 2003;290(16): [20]. Shinn EH, Poston WSC, Kimball KT, St. Jeor ST, Foreyt JP. Blood pressure and symptoms of depression and anxiety: a prospective study. American Journal of Hypertension. 2001;14(7): [21]. Shinagawa M, Otsuka K, Murakami S, et al. Seven-day (24-h) ambulatory blood pressure monitoring, self-reported depression and quality of life scores. Blood Pressure Monitoring. 2002;7(1): [22]. Okajima K, Yamanaka G, Oinuma S, et al. Even mild depression is associated with among-day blood pressure variability, including masked non-dipping assessed by 7-d/24-h ambulatory blood pressure monitoring. Clinical and Experimental Hypertension. 2015;37(5): [23]. Stroup-Benham CA, Markides KS, Black SA, Goodwin JS. Relationship between low blood pressure and depressive symptomatology in older people. Journal of the American Geriatrics Society. 2000;48(3): [24]. Barrett-Connor E, Palinkas LA. Low blood pressure and depression in older men: a population based study. British Medical Journal. 1994;308(6926): [25]. Newnham JP, Evans SF, Michael CA, Stanley FJ, Landau LI. Effects of frequent ultrasound during pregnancy: a randomised controlled trial. The Lancet. 1993;342(8876): [26]. Luppino FS, de Wit LM, Bouvy PF, et al. Overweight, obesity, and depression: A systematic review and meta-analysis of longitudinal studies. Archives of General Psychiatry. 2010;67(3):

143 [27]. Theodore RF, Broadbent J, Nagin D, et al. Childhood to early-midlife systolic blood pressure trajectories: early-life predictors, effect modifiers, and adult cardiovascular outcomes. Hypertension. 2015;66(6): [28]. Lovibond SH, Lovibond PF. (1995). Manual for the Depression Anxiety Stress Scales. (2nd Ed) Sydney: Psychology Foundation. [29]. Henry JD, Crawford JR. The short-form version of the Depression Anxiety Stress Scales (DASS-21): construct validity and normative data in a large non-clinical sample. British Journal of Clinical Psychology. 2005;44(2): [30]. Brunner Huber LR. Validity of self-reported height and weight in women of reproductive age. Maternal and Child Health Journal. 2007;11(2): [31]. Bhat SK BL, Robinson M, Burrows S, Mori T. Contrasting effects of prenatal life stress on blood pressure and BMI in young adults. Journal of Hypertension. 2015;33(4): [32]. Epstein NB, Baldwin LM, Bishop DS. The McMaster Family Assessment Device. Journal of Marital and Family Therapy. 1983;9(2): [33]. Jurakić D, Pedišić Ž, Andrijašević M. Physical activity of Croatian population: cross-sectional study using International Physical Activity Questionnaire. Croatian Medical Journal. 2009;50(2): [34]. Friedman MA, Brownell KD. Psychological correlates of obesity: Moving to the next research generation. Psychological Bulletin. 1995;117(1):3-20. [35]. Huang R-C, Burrows S, Mori TA, Oddy WH, Beilin LJ. Lifecourse adiposity and blood pressure between birth and 17 years old. American Journal of Hypertension. 2015;28(8): [36]. Shah MT, Zonderman AB, Waldstein SR. Sex and age differences in the relation of depressive symptoms with blood pressure. American Journal of Hypertension. 2013;26(12): [37]. Nabi H, Chastang J-F, Lefèvre T, et al. Trajectories of depressive episodes and hypertension over 24 years: The Whitehall II prospective cohort study. Hypertension. 2011;57(4):

144 [38]. Simon GE, Von Korff M, Saunders K, et al. Association between obesity and psychiatric disorders in the US adult population. Archives of General Psychiatry. 2006;63(7): [39]. Burke V, Beilin LJ, Dunbar D. Tracking of blood pressure in Australian children. Journal of Hypertension. 2001;19(7): [40]. Huffman JC, Celano CM, Beach SR, Motiwala SR, Januzzi JL. Depression and cardiac disease: epidemiology, mechanisms, and diagnosis. Cardiovascular Psychiatry and Neurology. 2013;2013:14. [41]. Ojike N, Sowers JR, Seixas A, et al. Psychological distress and hypertension: results from the National Health Interview Survey for Cardiorenal Medicine. 2016;6(3): [42]. 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: [43]. Barton DA, Dawood T, Lambert EA, et al. Sympathetic activity in major depressive disorder: identifying those at increased cardiac risk? Journal of Hypertension. 2007;25(10): [44]. Hu MX, Lamers F, de Geus EJC, Penninx BWJH. Differential autonomic nervous system reactivity in depression and anxiety during stress depending on type of stressor. Psychosomatic Medicine. 2016;78(5): [45]. Lu X-Y. The leptin hypothesis of depression: a potential link between mood disorders and obesity? Current Opinion in Pharmacology. 2007;7(6):

145 Chapter 6 Overall Discussion 6.1 Key findings in the context of the literature Emerging adulthood is a transitional period 286 during which behavioural patterns developed are likely to last for life and may be resistant to change 313. This is a period in which young adults often adopt adverse health behaviours such as smoking, excess alcohol intake, or consuming an unhealthy diet, that potentially affects cardiovascular well-being in later adulthood 313. Thus, emerging adulthood makes age 20 an especially rich, complex, dynamic period of life to study. The first study in this thesis investigated the effect of prenatal life stress score on offspring BP and BMI, at age 20. The most recent literature for the relationship between prenatal distress and offspring BP comes from three pregnancy cohort studies (ABCD, Generation-R, and ALSPAC studies). The Raine Study is to my knowledge the first pregnancy cohort study to examine the effect of prenatal life stress score on young adult BP and BMI, at age 20. Both cross-sectional and longitudinal association estimates were determined. While a higher prenatal life stress score was associated with higher BMI, an unexpected inverse relationship between prenatal life stress score and resting BP was also observed. The cross-sectional findings were similar in direction to the longitudinal estimates. Longitudinal analyses additionally adjusted for the correlation expected within the participants for their repeated measures of BP and BMI across the follow-up years. The longitudinal results, showed a significant positive effect of prenatal life stress score on offspring BMI from age 8 onwards and a significant inverse effect of prenatal life stress score on offspring systolic BP from age 14 onwards. The magnitude of the crosssectional estimation parameter estimate was nearly similar to the longitudinal estimation, an analytical technique that adjusts for participant s BP correlation over follow-up time. The cross-sectional results provided evidence for the dose-response relationship between multiple prenatal stress and BP at age 20. Thus, a greater effect was observed for the relationship between multiple prenatal stresses of 3 (compared to prenatal stress <3) and prehypertension/hypertension. This was in addition to the significantly inverse relationship between prenatal life stress score and BP. A similar significant inverse relationship in the ALSPAC study 280 at age 11 was restricted to 131

146 diastolic BP, but that study did not examine the BP relationship with multiple stressors. Additionally, the ABCD 278, Generation-R 279, and ALSPAC 280 studies did not examine the relationship between maternal stressors and offspring BMI. The measurement of stressful life events poses many challenges for researchers in medicine and the social sciences As exposure to multiple stressors during pregnancy not only is likely to be common but also subjective in that a stressful event for one person may not be stressful to another. However, the prenatal life stress events listed in this study are ranked high in the possible list of life events 317 and other studies that have weighted stressful life events according to their impact do not find it changes the results 318. The inverse association between BP and prenatal life stress was unexpected and the mechanisms are unclear. Several possible mechanisms derived from experimental animal models were presented in Chapter 3. It is plausible that prenatal stress may not be dramatically altering a given foetal structure or function, but affects resilience and renders the animal more susceptible to pathophysiological outcomes when further insults occur during adulthood 319. Young adult Wistar rats prenatally treated with the synthetic glucocorticoid dexamethasone (stress hormone) also displayed a lower basal BP in adulthood, despite a significantly greater rise in BP in response to stress 320. The second study examined the complex underlying influence of adverse family determinants on the offspring depression and adiposity relationship. This is important given the fact that lifestyle changes have proven to be feasible and beneficial for preventing depressed persons propensity to increased CVD risk 321. Maternal prenatal smoking is not only associated with a higher risk for depression diagnosis later in life 322,323 but overall health risks for both mother and offspring. This presents a window of opportunity for changing health behaviours in pregnancy , especially when many mothers are highly motivated to change for the benefit of their coming child 329. There is strong evidence of obesity linked to socioeconomic disadvantage in children and adults and smoking behaviours generally are linked more closely with low income, younger maternal age, and lower maternal education 219. Specifically, therefore this study examined the association between prevalent depression symptoms score and adiposity in young adults, and assessed the influence of maternal prenatal smoking or 132

147 family income on this association. The accentuating influence of depression symptoms with prenatal psychosocial or familial determinants was of special interest in an otherwise positive association between depression scores and adiposity. The offspring of mothers with a low family income at pregnancy (substituting for maternal prenatal smoking) also showed a positive association between depression score and BMI. Thus, adverse familial determinants such as maternal prenatal smoking and low family income adversely influenced the relationship between offspring depression score and adiposity. Socioeconomic disadvantage begets an increased risk of persistent depression in an individual 330 and individual s psychological and emotional distress may promote obesity 99. Given that maternal prenatal smoking is a proximal risk factor that reflects the broader influence of socioeconomic status on mental health outcomes 331 family income if substituted for maternal prenatal smoking should demonstrate influence similar to that of prenatal smoking on the relationship between offspring depression and adiposity. Indeed, the results of the second study showed the relationship between depression symptoms at age 20 and young adult BMI is seen predominantly in those with a history of prenatal smoking and low family income, highlighting lower socio-economic groups most suitable for targeting preventative health measures. The Raine study annual family income variable allowed to examine the influence of parental socioeconomic status by grouping family income above and under poverty line. A poverty line cut-off for the annual income is a practical marker of parental socioeconomic status unlike maternal education specific to mother. The relationship between anxiety scores and body mass index was not examined in the second study given the contradictory findings from a recent population based study using clinical data that showed non-linear U-shaped association between anxiety and body mass index categories among women 332. The third study examined the relationship between depression/anxiety symptoms and BP. This is to my knowledge the first study to show an inverse association between depression or anxiety symptoms score on resting SBP, at this stage of early adulthood age 20. The significant interaction between self-reported depression with offspring adiposity was of special interest. The interaction with adiposity was such that the inverse association between self-reported depression and systolic BP was accentuated with increasing BMI. SBP was expected to be associated positively with self-reported depression at least for those with higher BMI, rather than the directly opposite finding. 133

148 Mechanisms for the inverse relationship between SBP and depression are unclear. Several possible mechanisms were discussed in Chapter 5. Most of the existing literature has focussed on examining the relationship between depression and BP among elderly or middle-aged adults and have reported varying results, both in magnitude and direction. This study examined the relationship between depression symptoms and BP at emerging adulthood. Compared to older populations, the cusp of adulthood is a period when young adults are relatively free of physical health comorbidities , a source of major confounding in the adult studies examining CVD outcomes. The inverse association between depression or anxiety scores and systolic BP agrees with some of the cross-sectional and prospective studies in older populations 171,176. The Norwegian Hunt study reported an inverse association between depression and anxiety scores 169,185,186 in both men and women approximately one decade older than our cohort. The mechanisms for the association between depression and BP are not clear, though a positive relationship between depression and hypertension is biologically plausible because of autonomic imbalance and activation of the HPA axis 190,191. However, dysregulation does not always present with high cortisol levels. It may present with a blunted amplitude of cortisol secretion and impaired responsiveness to acute stressors 333. The Norwegian Hunt studies 169,185,186 not only had their cross-sectional findings validated with a more robust longitudinal study design but had a broad representation of ages extending from young adulthood to elderly. Their youngest age-group of years was closest to the Raine Study participants and was a chance to reconcile the findings between two geographically distinct cohorts. Whereas the DASS-21 subscales can validly be used to measure the dimensions of depression, anxiety, and stress each of these subscales also taps a more general dimension of psychological distress or negative affectivity 80. The logic of a shared variance underlying the Depression Anxiety Stress Scales if extended to prenatal stress, even though a different inventory, may allow postulation that the inverse direction of the association between prenatal stress score and BP mirrors the direction of the relationship between depression/anxiety score and BP, or vice-versa. 134

149 6.2 Limitations and strengths A limitation common to all the three studies was the loss of participants to follow-up, though all longitudinal studies are prone to attrition bias 334. The Raine Study attrition rate was 46.8% at the 20-year survey which is, similar to other birth cohorts over a long length of follow-up Given the proportion of loss to follow-up, the young adults surveyed, at age 20 may have differed from the original participants at recruitment. The study cohort having a greater retention of socially advantaged families would rather underestimate the observed influences, given obesity is more common in people from a lower socioeconomic background. Another, limitation is the cross-sectional design of the second and third studies that cannot prove cause and effect. Although longitudinal studies do not prove cause and effect relationships they benefit from additional adjustment of covariance between repeated measures. In the second study, the interaction effect of maternal prenatal smoking on the relation between offspring depression and adiposity is prospective in time. A general limitation, common to all observational studies, is a probable under-adjustment for confounders or unrecognised information 168. I was unable to examine the influence of SES (income) of the young adult at age 20 due to missing data for young adult income. The ascertainment of depression scores by questionnaire without formal clinical assessment of depression, clinic BPs on first visit and no ambulatory or home BPs may be considered a limitation to some extent. However, multiple follow-up BP readings with an average estimated BP can be considered appropriate in population studies 338. Despite the lack of clinical depression or BP assessment significant effects were found in the three studies. A major strength of the studies is the use of a large pregnancy cohort that provides detailed data collected from 18 weeks gestation, allowing prospective adjustments of a range of phenotypic, psychosocial, and behavioural risk factors from childhood through adolescence to young adulthood. Mixed effects modelling for BP and BMI in the first study not only used all longitudinal data but allowed a correction for within-subject correlations of their repeated measures for BP and BMI. Longitudinal analyses also helped validate the direction of cross-sectional estimates for prenatal life stress and BP/BMI relationship. The study on the relationship between depression scores and blood pressure in young adults has considered the interaction effect of depression and gender on the blood pressure outcome of young adults. Contrarily, not many authors or studies have looked into the gender differences in the relationship between depression 135

150 and BP in younger populations. In line with previous studies in adults that examined the association between depression or depression symptom scores and BMI or BP, most associations reported in all the three-studies survived multivariable adjustments for phenotypic, psychosocial, familial, and lifestyle confounders. The important confounders for the primary relationships were maternal prenatal smoking, female hormonal contraceptive use, maternal age at pregnancy and pre-pregnancy BMI for the BMI outcome in the second study; and female hormonal contraceptive use, offspring BMI and current alcohol consumption, birth weight and maternal hypertension in pregnancy for the BP outcome in the third study. The strength of the statistical analysis is of modelling estimates derived from an interaction equation, that provide a better fit in context of the research question. Analyses that rely on BMI categories substantially reduce the statistical power to detect the true association between stress events and later BP. Use of blood pressure categories instead of BP measurements as a continuum over the entire range is an inefficient use of available data. Continuous scores were preferred over categorical constructs because score continuum is more robust and meaningful for the associations. For instance, individuals who fall just short of a delineated clinical cutoff are correctly recognized as experiencing considerable symptoms and as being at high risk of developing more extreme symptoms 78. Thus, a strength of the three studies are the statistical estimates derived from regression of the full range of population BPs on prenatal life stress and depression or anxiety continuous scores in the first and third studies, respectively; and regression of continuous BMI on the prenatal life stress scores and depression scores in the first and second studies, respectively. 6.3 Conclusions and future research The studies presented in this thesis provide novel data demonstrating the influence of antenatal and postnatal psychosocial and familial determinants that underlie the relationships between young adulthood BP or BMI and prenatal life stress exposure and prevalent depression or anxiety scores at young adulthood. The first study showed contrasting effects of prenatal life stress exposure on BP and BMI in young adults. The second study showed an underlying role of adverse psychosocial and familial determinants (maternal prenatal smoking/ low family income at pregnancy) in the 136

151 positive association between depression symptoms and adiposity at age 20. The third study findings demonstrated systolic BP in young adults is inversely associated with depression or anxiety scores and self-reported depression attenuates the effect of BMI on BP or adiposity accentuates the inverse relationship between depression and SBP. In the first study effect modifications of BP, elicited by an interaction between prenatal stress score and BMI was a novelty. Similarly effect modification of BMI, elicited by the interaction between depression score and maternal prenatal smoking or family income was distinctive. In the third study, the influence of self-reported depression and BMI interaction on SBP appears to be unique in terms of the age group studied and counter to reports showing a predisposition of older depressed subjects for hypertension and CVD. The inverse associations between BP and prenatal life stresses and offspring depression or anxiety symptoms, are possibly due to underlying complex neuroendocrine mechanisms, but this needs to be clarified in future studies. The findings may help to identify vulnerable populations, thereby, aiding in population specific prevention strategies and intervention programs for reducing the life course CVD burden associated with co-morbid depression and obesity. The findings highlight the role of prenatal stresses, offspring negative emotions and their interactions with psychosocial, phenotypic, and family determinants in young adult BP or BMI outcomes. If the associations observed are causal, the findings suggest that unhealthy lifestyle and adverse health-related behaviours including their interactions are important antecedents of CVD in later life. In future, it would be valuable to assess longitudinal changes in offspring depression or anxiety in relation to BP and BMI in the Raine Study participants, now that the cohort is approaching 30 years of age. Modern methods applying genetics (Mendelian randomization) could shed light on these associations, for example the genetics of blood pressure regulation in relation to obesity and early life influences. 137

152 Appendix A Questionnaires A.1 Prenatal Life Stress Inventory at 18-week Gestation 138

153 A.1.1 Prenatal Life Stress Inventory at 34-week Gestation 139

154 A.2 DASS Inventory at Age

155 Appendix B Publication(s) B.1 Contrasting Effects of Prenatal Life Stress on BP and BMI in Young Adults 141

156 142

157 143

158 144

159 145

160 146

161 147

162 148

163 149

164 150

165 151