Association of Early-Life Mental Health With Biomarkers in Midlife and Premature Mortality: Evidence From the 1958 British Birth Cohort | Cardiology | JAMA Psychiatry | JAMA Network
[Skip to Navigation]
Sign In
Figure 1.  Ordinary Least Squares Regression Coefficients and 95% CIs of the Association Between Early-Life Mental Health and Midlife High-Density Lipoprotein (HDL) Cholesterol
Ordinary Least Squares Regression Coefficients and 95% CIs of the Association Between Early-Life Mental Health and Midlife High-Density Lipoprotein (HDL) Cholesterol

The stable low group is the reference group. Parameters are adjusted for birth weight, maternal smoking during pregnancy, maternal age, ever being breastfed, maternal partnership status at birth, maternal employment up to age 5 years, parents reading to child at age 7 years, parental interest in school at age 7 years, parents divorce by age 7 years, separation from child from more than 1 month at age 7 years, paternal social class at birth, financial difficulties at age 7 years, age mother stayed at school, housing tenure at age 7 years, access to household amenities at age 7 years, housing difficulties at age 7 years, cognitive ability at age 7 years, enuresis at age 7 years, a summary of health conditions assessed during the medical visit at age 7 years, region at birth, experience of bullying at age 7 years, and body mass index at age 7 years.

Figure 2.  Risk Ratios (RRs) and 95% CIs of the Association Between Early-Life Mental Health and Midlife Abdominal Obesity
Risk Ratios (RRs) and 95% CIs of the Association Between Early-Life Mental Health and Midlife Abdominal Obesity

The stable-low group is the reference group. Parameters are adjusted for birth weight, maternal smoking during pregnancy, maternal age, ever being breastfed, maternal partnership status at birth, maternal employment up to age 5 years, parents reading to child at age 7 years, parental interest in school at age 7 years, parents divorce by age 7 years, separation from child from more than 1 month at age 7 years, paternal social class at birth, financial difficulties at age 7 years, age mother stayed at school, housing tenure at age 7 years, access to household amenities at age 7 years, housing difficulties at age 7 years, cognitive ability at age 7 years, enuresis at age 7 years, a summary of health conditions assessed during the medical visit at age 7 years, region at birth, experience of bullying at age 7 years, and body mass index at age 7 years.

Table 1.  Latent Trait Scores of Conduct Problems and Affective Symptoms
Latent Trait Scores of Conduct Problems and Affective Symptoms
Table 2.  Descriptive Statistics of All Outcomesa
Descriptive Statistics of All Outcomesa
Table 3.  Association Between Early-Life Mental Health Trajectories, Biomarkers in Midlife, and Premature Mortalitya
Association Between Early-Life Mental Health Trajectories, Biomarkers in Midlife, and Premature Mortalitya
Supplement.

eTable 1. Descriptive statistics of early life mental health questions

eTable 2. Sample characteristics - Descriptive statistics of all potential confounders included in the model

eTable 3. Descriptive statistics of all continuous biomarkers in SI units

eTable 4. Latent Profile Analysis - Fit statistics

eTable 5. Modal allocation and BCH estimated means of fibrinogen and HbA1c

eTable 6. “Stable Low” sample characteristics - Descriptive statistics of all potential confounders included in the model

eTable 7. “Teacher Identified Adolescent Onset” sample characteristics - Descriptive statistics of all potential confounders included in the model

eTable 8. “Moderate” sample characteristics - Descriptive statistics of all potential confounders included in the model

eTable 9. “Stable High” sample characteristics - Descriptive statistics of all potential confounders included in the model

eTable 10. Latent Profile Analysis of maternal reports - Fit statistics

eTable 11. Joint distribution of trajectories based on maternal reports with trajectories with trajectories derived from maternal and teacher reports

eTable 12. Linear regression coefficients (fibrinogen, CRP, HDL, LDL, HbA1c, FEV1)/risk ratios (abdominal obesity & high blood pressure)/hazard ratios (all mortality outcomes) and 95% confidence intervals of the association between early life mental health trajectories (maternal reports only), biomarkers in midlife and premature mortality

eTable 13. Latent Profile Analysis of maternal reports for age 7 and 11 - Fit statistics

eTable 14. Joint distribution of trajectories based on maternal reports (ages 7 and 11) with trajectories with trajectories derived from maternal and teacher reports

eTable 15. Linear regression coefficients (fibrinogen, CRP, HDL, LDL, HbA1c, FEV1)/risk ratios (abdominal obesity & high blood pressure)/hazard ratios (all mortality outcomes) and 95% confidence intervals of the association between early life mental health trajectories (maternal reports, ages 7 & 11 only), biomarkers in midlife and premature mortality

eAppendix 1. Potential confounders

eAppendix 2. Sources of mortality data and classification system for cause specific mortality

eAppendix 3. Corrections for medications use

eAppendix 4. Latent Profile Analysis

eAppendix 5. Selection into the four early life mental health trajectories

eAppendix 6. Missing Data

eAppendix 7. Hazard ratios and 95% Cis of the association between early life mental health trajectories and suicide risk

eAppendix 8. Sensitivity analysis for unmeasured confounding

eAppendix 9. Latent Profile Analysis of maternal reports

eAppendix 10. Latent Profile Analysis of maternal reports only, ages 7 and 11

eAppendix 11. Sensitivity analysis with continuous measures of abdominal obesity and blood pressure

eFigure 1. Point estimates (risk ratios) and 95% CIs of confounders and the E – Value of the association between early life mental health and abdominal obesity (females)

eFigure 2. Abdominal obesity E Value plot

eFigure 3. Point estimates (hazard ratios) and 95% CIs of confounders and the E – Value of the association between early life mental health and all – cause mortality

eFigure 4. All – cause mortality E Value plot

eFigure 5. Regression coefficients and 95% confidence intervals of the association between early life mental health and blood pressure in midlife

eFigure 6. Regression coefficients and 95% confidence intervals of the association between early life mental health and waist-hip ratio in midlife

eReferences

1.
Clark  C, Rodgers  B, Caldwell  T, Power  C, Stansfeld  S.  Childhood and adulthood psychological ill health as predictors of midlife affective and anxiety disorders: the 1958 British Birth Cohort.   Arch Gen Psychiatry. 2007;64(6):668-678. doi:10.1001/archpsyc.64.6.668 PubMedGoogle ScholarCrossref
2.
Colman  I, Murray  J, Abbott  RA,  et al.  Outcomes of conduct problems in adolescence: 40 year follow-up of national cohort.   BMJ. 2009;338:a2981. doi:10.1136/bmj.a2981 PubMedGoogle ScholarCrossref
3.
Richards  M, Abbott  R. Childhood mental health and adult life chances in post-war Britain: insights from three national birth cohort studies. Smith Institute. Accessed August 24, 2020. http://www.smith-institute.org.uk/wp-content/uploads/2015/10/Childhoodmentalhealthandlifechancesinpostwarbritain.pdf
4.
Maughan  B, Stafford  M, Shah  I, Kuh  D.  Adolescent conduct problems and premature mortality: follow-up to age 65 years in a national birth cohort.   Psychol Med. 2014;44(5):1077-1086. doi:10.1017/S0033291713001402 PubMedGoogle ScholarCrossref
5.
Jokela  M, Ferrie  J, Kivimäki  M.  Childhood problem behaviors and death by midlife: the British National Child Development Study.   J Am Acad Child Adolesc Psychiatry. 2009;48(1):19-24. doi:10.1097/CHI.0b013e31818b1c76 PubMedGoogle ScholarCrossref
6.
Bardone  AM, Moffitt  TE, Caspi  A, Dickson  N, Stanton  WR, Silva  PA.  Adult physical health outcomes of adolescent girls with conduct disorder, depression, and anxiety.   J Am Acad Child Adolesc Psychiatry. 1998;37(6):594-601. doi:10.1097/00004583-199806000-00009 PubMedGoogle ScholarCrossref
7.
Winning  A, Glymour  MM, McCormick  MC, Gilsanz  P, Kubzansky  LD.  Psychological distress across the life course and cardiometabolic risk: findings from the 1958 British Birth Cohort Study.   J Am Coll Cardiol. 2015;66(14):1577-1586. doi:10.1016/j.jacc.2015.08.021 PubMedGoogle ScholarCrossref
8.
Patalay  P, Moulton  V, Goodman  A, Ploubidis  GB.  Cross-domain symptom development typologies and their antecedents: results from the UK Millennium Cohort Study.   J Am Acad Child Adolesc Psychiatry. 2017;56(9):765-776.e2. doi:10.1016/j.jaac.2017.06.009 PubMedGoogle ScholarCrossref
9.
Pinto Pereira  SM, Li  L, Power  C.  Early life factors and adult leisure time physical inactivity stability and change.   Med Sci Sports Exerc. 2015;47(9):1841-1848. doi:10.1249/MSS.0000000000000610 PubMedGoogle ScholarCrossref
10.
Maggs  JL, Patrick  ME, Feinstein  L.  Childhood and adolescent predictors of alcohol use and problems in adolescence and adulthood in the National Child Development Study.   Addiction. 2008;103(suppl 1):7-22. doi:10.1111/j.1360-0443.2008.02173.x PubMedGoogle ScholarCrossref
11.
Odgers  CL, Moffitt  TE, Broadbent  JM,  et al.  Female and male antisocial trajectories: from childhood origins to adult outcomes.   Dev Psychopathol. 2008;20(2):673-716. doi:10.1017/S0954579408000333 PubMedGoogle ScholarCrossref
12.
Goodman  A, Joyce  R, Smith  JP.  The long shadow cast by childhood physical and mental problems on adult life.   Proc Natl Acad Sci U S A. 2011;108(15):6032-6037. doi:10.1073/pnas.1016970108 PubMedGoogle ScholarCrossref
13.
Angold  A.  Childhood psychopathology can be really bad for your health.   J Am Acad Child Adolesc Psychiatry. 2009;48(1):3-4. doi:10.1097/CHI.0b013e3181908c49 PubMedGoogle ScholarCrossref
14.
Power  C, Elliott  J.  Cohort profile: 1958 British birth cohort (National Child Development Study).   Int J Epidemiol. 2006;35(1):34-41. doi:10.1093/ije/dyi183 PubMedGoogle ScholarCrossref
15.
Rutter  M, Tizard  J, Whitmore  K.  Education, Health and Behaviour. Longman Publishing Group; 1970.
16.
De Los Reyes  A.  Introduction to the special section: More than measurement error: Discovering meaning behind informant discrepancies in clinical assessments of children and adolescents.   J Clin Child Adolesc Psychol. 2011;40(1):1-9. doi:10.1080/15374416.2011.533405PubMedGoogle ScholarCrossref
17.
Van Roy  B, Groholt  B, Heyerdahl  S, Clench-Aas  J.  Understanding discrepancies in parent-child reporting of emotional and behavioural problems: effects of relational and socio-demographic factors.   BMC Psychiatry. 2010;10:56-56. doi:10.1186/1471-244X-10-56 PubMedGoogle ScholarCrossref
18.
Yoon  YS, Oh  SW.  Optimal waist circumference cutoff values for the diagnosis of abdominal obesity in Korean adults.   Endocrinol Metab (Seoul). 2014;29(4):418-426. doi:10.3803/EnM.2014.29.4.418 PubMedGoogle ScholarCrossref
19.
Lean  MEJ, Han  TS, Morrison  CE.  Waist circumference as a measure for indicating need for weight management.   BMJ. 1995;311(6998):158-161. doi:10.1136/bmj.311.6998.158 PubMedGoogle ScholarCrossref
20.
Fuller  E, Power  C, Shepherd  P, Strachan  D. Technical report on the National Child Development Study biomedical survey 2002-2004. National Centre for Social Research. Published July 2006. https://cls.ucl.ac.uk/wp-content/uploads/2017/07/NCDS-biomed-technical-report.pdf
21.
Cole  TJ, Altman  DG.  Statistics Notes: Percentage differences, symmetry, and natural logarithms.   BMJ. 2017;358:j3683. doi:10.1136/bmj.j3683PubMedGoogle ScholarCrossref
22.
Pinto Pereira  SM, Ki  M, Power  C.  Sedentary behaviour and biomarkers for cardiovascular disease and diabetes in mid-life: the role of television-viewing and sitting at work.   PLoS One. 2012;7(2):e31132. doi:10.1371/journal.pone.0031132 PubMedGoogle Scholar
23.
Ki  M, Pouliou  T, Li  L, Power  C.  Physical (in)activity over 20 y in adulthood: associations with adult lipid levels in the 1958 British birth cohort.   Atherosclerosis. 2011;219(1):361-367. doi:10.1016/j.atherosclerosis.2011.07.109 PubMedGoogle ScholarCrossref
24.
Geoffroy  MC, Gunnell  D, Power  C.  Prenatal and childhood antecedents of suicide: 50-year follow-up of the 1958 British Birth Cohort study.   Psychol Med. 2014;44(6):1245-1256. doi:10.1017/S003329171300189X PubMedGoogle ScholarCrossref
25.
Case  A, Deaton  A.  Rising morbidity and mortality in midlife among white non-Hispanic Americans in the 21st century.   Proc Natl Acad Sci U S A. 2015;112(49):15078-15083. doi:10.1073/pnas.1518393112 PubMedGoogle ScholarCrossref
26.
Muthén  B.  A general structural equation model with dichotomous, ordered categorical, and continuous latent variable indicators.   Psychometrika. 1984;49(1):115-132. doi:10.1007/BF02294210 Google ScholarCrossref
27.
Rabe-Hesketh  S, Skrondal  A.  Classical latent variable models for medical research.   Stat Methods Med Res. 2008;17(1):5-32. doi:10.1177/0962280207081236 PubMedGoogle ScholarCrossref
28.
Oberski  D. Mixture models: Latent profile and latent class analysis. In: Robertson J, Kaptein M, eds.  Modern Statistical Methods for HCI. Springer; 2016:275-287. doi:10.1007/978-3-319-26633-6_12
29.
Pang  M, Kaufman  JS, Platt  RW.  Studying noncollapsibility of the odds ratio with marginal structural and logistic regression models.   Stat Methods Med Res. 2016;25(5):1925-1937. doi:10.1177/0962280213505804 PubMedGoogle ScholarCrossref
30.
White  IR, Royston  P, Wood  AM.  Multiple imputation using chained equations: issues and guidance for practice.   Stat Med. 2011;30(4):377-399. doi:10.1002/sim.4067 PubMedGoogle ScholarCrossref
31.
Frane  AV.  Are per-family type I error rates relevant in social and behavioral science?   J Mod Appl Stat Methods. 2015;14(1):5. doi:10.22237/jmasm/1430453040Google ScholarCrossref
32.
Archer  G, Kuh  D, Hotopf  M, Stafford  M, Richards  M.  Adolescent affective symptoms and mortality.   Br J Psychiatry. 2018;213(1):419-424. doi:10.1192/bjp.2018.90PubMedGoogle ScholarCrossref
33.
Moffitt  TE, Caspi  A.  Psychiatry’s opportunity to prevent the rising burden of age-related disease.   JAMA Psychiatry. 2019;76(5):461-462. doi:10.1001/jamapsychiatry.2019.0037 PubMedGoogle ScholarCrossref
34.
Ploubidis  GB, Sullivan  A, Brown  M, Goodman  A.  Psychological distress in mid-life: evidence from the 1958 and 1970 British birth cohorts.   Psychol Med. 2017;47(2):291-303. doi:10.1017/S0033291716002464 PubMedGoogle ScholarCrossref
35.
Collishaw  S, Maughan  B, Natarajan  L, Pickles  A.  Trends in adolescent emotional problems in England: a comparison of two national cohorts twenty years apart.   J Child Psychol Psychiatry. 2010;51(8):885-894. doi:10.1111/j.1469-7610.2010.02252.x PubMedGoogle ScholarCrossref
36.
Patalay  P, Gage  SH.  Changes in millennial adolescent mental health and health-related behaviours over 10 years: a population cohort comparison study.   Int J Epidemiol. 2019;48(5):1650-1664. doi:10.1093/ije/dyz006 PubMedGoogle ScholarCrossref
37.
Joyce  R, Xu  X. Inequalities in the Twenty-First Century: Introducing the IFS Deaton Review. Institute for Fiscal Studies. Published May 2019. https://www.ifs.org.uk/inequality/wp-content/uploads/2019/05/The-IFS-Deaton-Review-launch.pdf
38.
Leon  DA, Jdanov  DA, Shkolnikov  VM.  Trends in life expectancy and age-specific mortality in England and Wales, 1970-2016, in comparison with a set of 22 high-income countries: an analysis of vital statistics data.   Lancet Public Health. 2019;4(11):e575-e582. doi:10.1016/S2468-2667(19)30177-X PubMedGoogle ScholarCrossref
39.
Gondek  D, Bann  D, Ning  K, Grundy  E, Ploubidis  GB.  Post-war (1946-2017) population health change in the United Kingdom: A systematic review.   PLoS One. 2019;14(7):e0218991. doi:10.1371/journal.pone.0218991 PubMedGoogle Scholar
40.
Birnie  K, Cooper  R, Martin  RM,  et al; HALCyon study team.  Childhood socioeconomic position and objectively measured physical capability levels in adulthood: a systematic review and meta-analysis.   PLoS One. 2011;6(1):e15564. doi:10.1371/journal.pone.0015564 PubMedGoogle Scholar
41.
Deary  IJ, Whiteman  MC, Starr  JM, Whalley  LJ, Fox  HC.  The impact of childhood intelligence on later life: following up the Scottish mental surveys of 1932 and 1947.   J Pers Soc Psychol. 2004;86(1):130-147. doi:10.1037/0022-3514.86.1.130 PubMedGoogle ScholarCrossref
42.
Lacey  RE, Pinto Pereira  SM, Li  L, Danese  A.  Adverse childhood experiences and adult inflammation: single adversity, cumulative risk and latent class approaches.   Brain Behav Immun. 2020;87:820-830. doi:10.1016/j.bbi.2020.03.017 PubMedGoogle ScholarCrossref
43.
Brooks  SK, Webster  RK, Smith  LE,  et al.  The psychological impact of quarantine and how to reduce it: rapid review of the evidence.   Lancet. 2020;395(10227):912-920. doi:10.1016/S0140-6736(20)30460-8 PubMedGoogle ScholarCrossref
44.
Galea  S, Merchant  RM, Lurie  N.  The mental health consequences of COVID-19 and physical distancing: the need for prevention and early intervention.   JAMA Intern Med. 2020. doi:10.1001/jamainternmed.2020.1562 PubMedGoogle Scholar
45.
Kendler  KS, Gardner  CO.  Sex differences in the pathways to major depression: a study of opposite-sex twin pairs.   Am J Psychiatry. 2014;171(4):426-435. doi:10.1176/appi.ajp.2013.13101375 PubMedGoogle ScholarCrossref
46.
Pongiglione  B, De Stavola  BL, Kuper  H, Ploubidis  GB.  Disability and all-cause mortality in the older population: evidence from the English Longitudinal Study of Ageing.   Eur J Epidemiol. 2016;31(8):735-746. doi:10.1007/s10654-016-0160-8 PubMedGoogle ScholarCrossref
47.
Verbrugge  LM.  Sex differentials in health.   Public Health Rep. 1982;97(5):417-437.PubMedGoogle Scholar
48.
Patton  GC, Coffey  C, Romaniuk  H,  et al.  The prognosis of common mental disorders in adolescents: a 14-year prospective cohort study.   Lancet. 2014;383(9926):1404-1411. doi:10.1016/S0140-6736(13)62116-9 PubMedGoogle ScholarCrossref
49.
VanderWeele  TJ, Ding  P.  Sensitivity analysis in observational research: introducing the E-value.   Ann Intern Med. 2017;167(4):268-274. doi:10.7326/M16-2607 PubMedGoogle ScholarCrossref
50.
Ding  P, VanderWeele  TJ.  Sensitivity analysis without assumptions.   Epidemiology. 2016;27(3):368-377. doi:10.1097/EDE.0000000000000457 PubMedGoogle ScholarCrossref
51.
Rose  G.  Sick individuals and sick populations.   Int J Epidemiol. 1985;14(1):32-38. doi:10.1093/ije/14.1.32 PubMedGoogle ScholarCrossref
Limit 200 characters
Limit 25 characters
Conflicts of Interest Disclosure

Identify all potential conflicts of interest that might be relevant to your comment.

Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.

Err on the side of full disclosure.

If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.

Not all submitted comments are published. Please see our commenting policy for details.

Limit 140 characters
Limit 3600 characters or approximately 600 words
    Original Investigation
    September 30, 2020

    Association of Early-Life Mental Health With Biomarkers in Midlife and Premature Mortality: Evidence From the 1958 British Birth Cohort

    Author Affiliations
    • 1Department of Epidemiology and Public Health, University College London, London, England
    • 2School of Biological and Population Health Sciences, Oregon State University, Corvallis
    • 3Centre for Longitudinal Studies, Department of Social Science, UCL Institute of Education, University College London, London, England
    • 4MRC Unit for Lifelong Health and Ageing, Department of Population Science and Experimental Medicine, University College London, London, England
    JAMA Psychiatry. 2021;78(1):38-46. doi:10.1001/jamapsychiatry.2020.2893
    Key Points

    Question  Is early-life mental health associated with biomarkers in midlife and premature mortality?

    Findings  In a birth cohort study of 17 415 participants, early-life mental health was associated with less favorable levels of biomarkers 28 years later and elevated risk of premature mortality. The associations with biomarkers in midlife were more commonly seen in female individuals.

    Meaning  These findings, if causal and generalizable to younger cohorts, may have implications for overall population health and age-related disease as effective interventions on early-life mental health have the potential to shift the distribution of risk and improve population health.

    Abstract

    Importance  Early-life mental health is known to be associated with socioeconomic adversity and psychological distress in adulthood, but less is known about potential associations with biomarkers and mortality.

    Objective  To investigate the association between early-life mental health trajectories with biomarkers in midlife and premature mortality.

    Design, Setting, and Participants  This study used data from the British National Child Development Study, a population-based birth cohort. The initial sample of 17 415 individuals consisted of all infants born in Great Britain in a single week in 1958. Analysis began Feburary 2017 and ended May 2020.

    Main Outcomes and Measures  Biomarkers collected at age 44 to 45 years were fibrinogen, C-reactive protein, glycated hemoglobin, high- and low-density lipoprotein cholesterol, forced expiratory volume, blood pressure, and waist-to-hip ratio. Information on all-cause mortality was available up to age 58 years and cause-specific mortality up to age 50 years.

    Results  Biomarkers were analyzed from 9377 participants (age, 44-45 years, 4712 female [50.3%]) and mortality data from 15 067 participants (age, 58 years; 7379 female [49.0%]). A 4-group longitudinal typology of early-life conduct problems and affective symptoms was identified: (1) stable low, (2) teacher-identified adolescent onset, (3) moderate, and (4) stable high. Compared with the stable-low group, the stable-high (B, 2.308; 95% CI, 0.213-4.404) and teacher-identified adolescent-onset (B, 2.114; 95% CI, 0.725-3.503) groups had less favorable levels of fibrinogen in middle age. Effect modification was observed by sex (P < .005) such that compared with the stable-low group, differences in high-density lipoprotein and abdominal obesity were only observed in female individuals. The stable-high and teacher-identified adolescent-onset groups had elevated risk for all-cause mortality (hazard ratio, 1.878; 95% CI, 1.501-2.350 and hazard ratio, 1.412; 95% CI, 1.174-1.698). Psychopathology-associated mortality was higher in both groups but unintentional injuries–associated mortality only in the stable-high group.

    Conclusions and Relevance  Experiencing affective symptoms and conduct problems in childhood and adolescence is associated with less favorable levels of biomarkers 28 years later and elevated risk of premature mortality. These findings, if causal and generalizable to younger cohorts, may imply that effective interventions on early-life mental health have the potential to shift the distribution of risk and improve population health.

    Introduction

    Mental health problems in childhood and adolescence are known to be associated with an array of unfavorable outcomes in later life, including psychological distress,1 low educational attainment, unemployment, unstable family formation, and criminal offending.2,3 There are also emerging links with premature all-cause mortality4,5 but a paucity of evidence concerning the prospective association with physical health in adulthood. In the few studies that have investigated this link,6,7 the developmental perspective in the emergence of mental health symptomatology in childhood and adolescence8 has been neglected.

    There are various plausible mechanisms of action through which early-life mental health may be linked with midlife health and premature mortality. For example, early-life mental health has been shown to be associated with lower levels of physical activity and a greater prevalence of smoking, alcohol use,9-11 and socioeconomic adversity in adulthood,3,11,12 while specific links to mortality include suicide, drug overdose, unintentional injuries, and homicide.13 We capitalize on the availability of 3 assessments of mental health spanning childhood and adolescence in a population-based prospective birth cohort to characterize trajectories of early-life affective symptoms and conduct problems and to investigate their association with biomarkers in midlife and premature mortality.

    Methods
    Data

    Participants in the British National Child Development Study14 have been resurveyed on 10 occasions since birth. The initial sample of 17 415 individuals, consisting of all infants born in Great Britain in a single week in 1958 provide prospective data on social, biological, physical, and psychological phenotypes. In 2002 and 2003, when cohort members were aged 44 to 45 years, a biomedical survey was conducted in 9377 respondents. Ethical approval for this study was obtained by the University College London Institute of Education Research Ethics Committee. Written consent for all measures and samples was collected during the interview.

    Measures
    Early-Life Mental Health

    Conduct problems and affective symptoms in childhood and adolescence were assessed using the modified version of the Rutter Child Scale A.15 This version of the scale was completed by mothers of participants at ages 7 and 11 years and by mothers and teachers at age 16 years. Mother and teacher reports were used to capture symptoms at home and school, as is well known that maternal and teacher reports are weakly correlated and triangulating information from multiple informants may bring unique insights from different environments into children’s behavior.16,17 More information on the Rutter Child Scale is available in eTable 1 and eAppendix 4 in the Supplement.

    Biomarkers in Midlife

    We used biomarkers as collected at age 44 to 45 years: fibrinogen, a marker of inflammation and cardiovascular disease; C-reactive protein (CRP), an indicator of inflammation and cardiovascular disease; glycated hemoglobin, an index of glucose metabolism throughout the previous 30 to 90 days, which is used as a marker of diabetes; high-density lipoprotein (HDL) and low-density lipoprotein cholesterol as markers of cardiovascular risk; and high blood pressure (3 measures of systolic and diastolic blood pressure were taken). The mean of valid readings was used, and an individual was recorded as having high blood pressure if the mean value was above 140/90 mm Hg. Participants who reported use of blood pressure–related medication were classified as having high blood pressure. Waist and hip circumferences were measured and the ratio of waist over hip was used as a measure of abdominal obesity (≥0.94 in male individuals and ≥0.85 in female individuals18,19). Results with continuous versions of both dichotomized, for comparison purposes with other studies, outcomes were similar to those presented here (eAppendix 11 and eFigures 5-6 in the Supplement). To assess respiratory function, we used forced expiratory volume, with the highest measurement used. Forced expiratory volume is a measure of how much air a person can exhale during forced breath during the first second. Details on the measurement procedures and equipment are available on the biomedical sweep technical report.20 With the exception of glycated hemoglobin, which represents the percentage of total hemoglobin, all other continuous biomarkers were transformed with the natural logarithm and multiplied by 100, so all ordinary least squares regression coefficients can be interpreted as percentage differences in means.21 Appropriate corrections for medication use were also applied22,23 (eAppendix 3 and eTable 3 in the Supplement).

    Mortality

    Mortality was recorded by National Health Service digital notifications combined with information during fieldwork and the address database held at the Centre for Longitudinal Studies (eAppendix 2 in the Supplement). All-cause mortality data were available up to age 58 years and cause-specific mortality up to age 50 years, both with information on year of death. We analyzed 3 causes of death that have been hypothesized to be associated with early-life mental health5,13,24: (1) unintentional injuries (transport-related injuries and other external causes of unintentional injury); (2) deaths due to drug and alcohol poisonings, suicide, chronic liver diseases, and cirrhosis, collectively known as death of despair25; and (3) deaths related to the circulatory system. Further information on the classification of deaths and results with suicide risk as a stand-alone outcome are available in eAppendix 3 and eAppendix 7 in the Supplement.

    Potential Confounders

    We included a rich set of early-life potential confounders such as birth characteristics, parental characteristics, indicators of early-life socioeconomic position, and characteristics of the cohort member from birth to age 7 years. Detailed information about potential confounders is presented in eAppendix 1 and eTable 2 in the Supplement.

    Statistical Modeling
    Analytic Strategy

    We derived separate latent summaries of conduct problems and affective symptoms at ages 7, 11, and 16 years by modeling the probability of response to the Rutter Child Scale items with a 2-parameter probit latent variable measurement model.26,27 The conduct problems and affective symptoms scores were entered in a latent profile analysis28 to derive a longitudinal typology of mental health from ages 7 to 16 years that was used as a predictor of biomarkers in midlife and mortality in models that included confounders up to age 7 years (eAppendix 4 in the Supplement). For continuous outcomes, we estimated ordinary least squares regression and for binary outcomes (abdominal obesity and high blood pressure), a log binomial model with robust standard errors that returns risk ratios as both outcomes were not rare (>20%) to avoid bias due to noncollapsibility of the odds ratio.29 The association between early-life mental health trajectories and mortality was analyzed with the standard Cox regression model. Considering that female individuals live longer than male individuals and that female individuals are on average less healthy, we investigated modification of all hypothesized associations by sex. We used Stata software versions 15 and 16 (StataCorp) and Mplus software, version 8 (Muthén and Muthén). Two-sided P values with varying levels of significance due to corrections for multiple testing were used. Analysis began in February 2017 and ended in May 2020.

    Missing Data

    We used multiple imputation with chained equations30 and generated 50 imputed data sets to reduce bias and maintain power. More information on our missing data analysis strategy is available in eAppendix 6 in the Supplement.

    Results
    Early-Life Mental Health Trajectories

    A 4-group solution was selected based on various criteria and quality of allocation/misclassification indices (eAppendix 4 and eTable 4 in the Supplement). Descriptive statistics of all summary variables are presented in Table 1. The 4 longitudinal groups included (1) the stable-low group (5779 [46.3%]), which was characterized by absence of conduct problems and affective symptoms at all ages; (2) the teacher-identified adolescent-onset (TIAO) group (2236 [17.9%]) (absence of affective symptoms and conduct problems at ages 7 and 11 years, followed by sharp increase in levels of both dimensions, but predominantly conduct problems, at age 16 years as reported by teachers); (3) the moderate group (3472 [27.8%]), which experienced moderate levels of conduct problems and affective symptoms in childhood (ages 7 and 11 years), followed by a decline, which signifies improvement, in both dimensions at age 16 years as reported by teachers but a mild increase in both dimensions reported by mothers; and (4) the stable-high group (994 [8%]), in which persistent experience of both conduct problems and affective symptoms from age 7 to 16 years were reported consistently by both mothers and teachers. More information on selection to the 4 groups is available in eAppendix 5 and eTables 6 to 9 in the Supplement.

    Biomarkers

    The analytic sample ranged from 7466 (CRP) to 9298 (abdominal obesity) individuals depending on the outcome. A systematic pattern emerged with respect to all outcomes (Table 2), with the 4 groups being consistently ordered. The stable-high group had the least favorable levels on all biomarkers and the highest mortality rates. The TIAO group had the second worst biomarker levels and mortality rates, followed by the moderate group. The stable-low group had the most favorable biomarker levels and the lowest mortality rates. In Table 3, we present ordinary least squares regression coefficients, risk ratios, and hazard ratios for the association between early-life mental health trajectories and all outcomes after multivariable adjustment. With a conventional 5% α level, early-life mental health was associated with less favorable levels of fibrinogen, CRP, HDL cholesterol, and abdominal obesity. The associations with fibrinogen, HDL, and abdominal obesity were owing to the stable-high group having different levels in these biomarkers compared with the stable-low group. Furthermore, compared with the stable-low group, the TIAO group had higher levels of fibrinogen and CRP and had increased risk for abdominal obesity.

    We observed evidence of modification by sex in the association between early-life mental health, HDL, and abdominal obesity. With respect to HDL, there was strong evidence of interaction with sex in the TIAO group (B for interaction, −6.203; 95% CI, −9.333 to −3.074; P < .001) and considering that interaction terms require more power, there was also suggestive evidence for the stable-high group (B for interaction, −4.214; 95% CI, −8.795 to 0.367; P = .07). For abdominal obesity, there was strong interaction with sex in the stable-high group (risk ratio for interaction, 1.378; 95% CI, 1.109-1.711; P = .004). In Figure 1 and Figure 2, we present results (point estimates and 95% CIs) from models stratified by sex, where early-life mental health is associated with both outcomes only in female individuals. With respect to HDL, the TIAO and stable-high groups had less favorable levels compared with the stable-low group (B, −3.404; 95% CI, −5.699 to −1.110; P = .004 and B, −3.986; 95% CI, −7.351 to −0.621; P = .02, respectively). Similarly, the TIAO and stable-high groups had increased risk for abdominal obesity, compared with the stable-low group (risk ratio, 1.190; 95% CI, 1.019-1.389; P = .03 and risk ratio, 1.522; 95% CI, 1.278-1.813; P < .001).

    From a null hypothesis significance testing perspective, carrying out multiple tests increases the probability of type I error. We further evaluated our findings with the Bonferroni correction31 that controls for familywise error rate. Considering each biomarker as a family of tests, with 3 tests carried out, the α level was .05 / 3 = .016. Using this level, we were able to detect associations with fibrinogen, CRP, HDL (female individuals only), and abdominal obesity. In a more conservative scenario, where all biomarkers were considered a family of tests, the Bonferroni-adjusted α level would be .05 / (3 × 8) = .002. With this very conservative correction, we were able to detect an association with abdominal obesity. Going beyond the conventional null hypothesis significance testing dichotomy and considering the magnitude and patterns of point estimates as well as coverage of CIs as presented in Table 3, we observed a systematic pattern of associations with fibrinogen, CRP, HDL, glycated hemoglobin, and abdominal obesity.

    Mortality

    Having excluded participants who had died prior to age 16 years, our analytic sample for the mortality analyses was n = 15 067. With no evidence of effect modification by sex in the mental health–mortality association, we present results from the pooled sample. For those who participated in the data collection at age 16 years, there were 1068 deaths and we analyzed a total of 765 209 person-years, with age 58 years being the last year of observation. More deaths were observed in the stable-high and TIAO groups compared with the moderate and stable-low groups (Table 2). The early-life mental health trajectories were associated with premature all-cause mortality, with stable-high and TIAO groups having elevated risk of mortality compared with the stable-low group (stable-high group: hazard ratio, 1.878; 95% CI, 1.501-2.350; P < .001; TIAO group: hazard ratio, 1.412; 95% CI, 1.174-1.698; P < .001). We were able to detect an association with all-cause mortality under any familywise error rate correction.

    With respect to cause-specific mortality, we analyzed a total of 718 254 person-years, with age 50 years being the last year of observation. For those who participated in the data collection at age 16 years, we observed 82 unintentional injuries, 82 deaths due to drug and alcohol poisonings, suicide, chronic liver diseases, and cirrhosis (with 55 suicides among those), and 154 circulatory system–related deaths. Considering the low statistical power owing to the small number of events, the stable-high group had elevated risk for unintentional injuries and deaths due to drug and alcohol poisonings, suicide, chronic liver diseases, and cirrhosis, as well as suicide when analyzed separately (results for suicide analyzed separately available in eAppendix 7 in the Supplement). The TIAO group had elevated risk for deaths due to drug and alcohol poisonings, suicide, chronic liver diseases, and cirrhosis. The elevated risk for these deaths in the stable-high compared with the stable-low group was detectable under any family wise error rate correction (hazard ratio, 3.167; 95% CI, 1.529-6.558; P = .002).

    Discussion

    Our main finding was that we observed associations between early-life mental health with biomarkers in midlife as well as premature mortality. The associations with biomarkers were more commonly seen in female individuals, whereas as expected from earlier follow-ups of the present birth cohort and 1 other study,4,5,24,32 associations with mortality were observed in both male and female individuals. Persistent experience of high levels of conduct problems and affective symptoms from age 7 to 16 years was associated with higher fibrinogen and CRP levels, lower HDL levels, and increased risk for abdominal obesity 28 years later. It was also associated with elevated risk for premature all-cause mortality and when separately examined, unintentional injuries and deaths due to drug and alcohol poisonings, suicide, chronic liver diseases, and cirrhosis. Experiencing the onset of high levels of conduct problems and affective symptoms in adolescence was associated with CRP and abdominal obesity in midlife, as well as increased risk for premature all-cause mortality and suicide (results for suicide risk available in eAppendix 7 in the Supplement). However, we did not observe an association with unintentional injuries nor cardiovascular mortality, which indicates that trajectory-specific mechanisms underlie the association with premature mortality. In this group, it appears to be primarily driven by direct results of psychopathology, whereas in the stable-high group was also driven by unintentional injuries and perhaps by circulatory system deaths, but our estimates were imprecise.

    Our findings, if causal and generalizable to younger cohorts, may have implications for overall population health but also for age-related disease as prevention of morbidity and disability in old age could be achieved by improving early-life mental health.33 The rising levels of early-life and midlife psychological distress in more recently born cohorts34-36 point to early-life mental health being one of the drivers of the increases in midlife mortality caused by deaths due to drug and alcohol poisonings, suicide, chronic liver diseases, and cirrhosis37 and perhaps to the recent stagnation of life expectancy in the UK.38 The observed link with markers of health and mortality implies that generational differences in early-life mental health may partly underlie the observed expansion of morbidity in the UK.39 As a target of interventions, early-life mental health has the potential to mitigate the effect that adverse socioeconomic circumstances in childhood, adverse childhood experiences, and early-life low cognitive ability have in adulthood as it can be plausibly assumed to lie on the pathway that links these with adult health, mortality, and social and economic outcomes.40-42 Furthermore, interventions to improve early-life mental health could prevent future excess morbidity and mortality due to the possible (at the time of writing) psychological effect on children and adolescents that coronavirus disease 2019–related measures might have.43,44

    Experiencing the onset of mental health symptoms in the transition from childhood to adolescence was found to be detrimental with respect to both biomarkers and premature mortality. As the assessment of individuals prior to 16 years was carried out at age 11 years, it appears that the transition from childhood to adolescence may be a sensitive period that warrants further research to elucidate the mechanisms that links the onset of mental health problems during this period to adult health and premature mortality.

    The mechanisms of action that underlie the observed modification by sex in the association with HDL and abdominal obesity that were only observed in female individuals may include societal pressures and/or sex-related inequalities that may exacerbate the association of conduct problems and affective symptoms with midlife health in female individuals, pointing to sex-specific pathways, in a similar manner as to those suggested for adult depression.45 It appears that at least in this cohort, early-life mental health may be one of the factors that underlie the so-called sex survival paradox, a term that describes the observation that female individuals are on average less healthy but live longer than male individuals.46,47

    Strengths and Limitations

    Strengths of this study include the availability of a large population-based and representative prospective study, the 3 assessments of early-life mental health, which is of major importance given the time-varying nature of mental health in childhood and adolescence,48 the inclusion of both conduct problems and affective symptoms in the assessment of early-life mental health, and the wealth of information on potential confounders. However, our findings can only be generalized to those born in Britain in 1958 or close to this year. Furthermore, our data are derived from an observational longitudinal study and bias due to unmeasured confounding cannot be ruled out. However, sensitivity analyses with the E value49,50 suggest that for the observed associations with abdominal obesity and all-cause mortality, strong confounding, stronger than that observed in our data, would be needed to completely explain away our findings (eAppendix 8 and eFigures 1-4 in the Supplement). An alternative specification of our models where early-life mental health trajectories where restricted to data from ages 7 and 11 years, returned similar results (eAppendix 10 and eTables 13-15 in the Supplement). This analysis considerably shortened the temporal gap from the measurement of the confounders to when early-life mental health was assessed and provided further reassurance against unmeasured confounding. As in any longitudinal survey, missing data due to attrition are unavoidable. We used multiple imputation, augmenting our models with auxiliary variables in the imputation phase to maximize the plausibility of the missing-at-random assumption and restore sample representativeness, but bias due to a nonignorable missing data-generating mechanism cannot be ruled out.

    We did not include hyperactivity symptoms in our analysis, since building on previous work, we focused on affective symptoms and conduct problems as potential mechanisms of action with our outcomes have been shown to exist in the National Child Development Study and other cohorts. Another limitation is that reports from both mothers and teachers were only available at age 16 years. While they agreed in their assessment of those who experienced persistent mental health symptoms (stable-high group) and those who did not (stable-low group), there was stark disagreement in the TIAO group. The similar assessments for 2 of the 4 groups makes the possibility of teachers misreporting unlikely. A more plausible explanation is that symptoms were context specific and lack of insight into the other context might lead to misreporting. However, results from sensitivity analysis using only maternal reports were similar to our main analysis, indicating that our findings are not reporter driven (eAppendix 9 and eTables 10-12 in the Supplement). Further sensitivity analysis taking into account measurement error in the mental health trajectories returned similar results to the ones presented here (eAppendix 4 and eTable 5 in the Supplement). Despite these efforts, the extent to which undetected systematic measurement error may have biased our results is unknown.

    Conclusions

    Our findings, if causal and generalizable to younger cohorts, may have implications for public health policy, especially if mental health is worse in more recently born cohorts.34,35 Effective interventions on early-life mental health have the potential to shift the distribution of risk51 and may have additional benefits, reducing future risk of premature mortality and improving multiple physical health outcomes.

    Back to top
    Article Information

    Corresponding Author: George B. Ploubidis, PhD, Centre for Longitudinal Studies Department of Social Science, University College London, 55-59 Gordon Square, Room 212, London WC1H 0NU, England (g.ploubidis@ucl.ac.uk).

    Accepted for Publication: July 19, 2020.

    Published Online: September 30, 2020. doi:10.1001/jamapsychiatry.2020.2893

    Author Contributions: Dr Ploubidis had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

    Concept and design: Ploubidis, Batty, Bann, Goodman.

    Acquisition, analysis, or interpretation of data: Ploubidis, Batty, Patalay, Goodman.

    Drafting of the manuscript: Ploubidis.

    Critical revision of the manuscript for important intellectual content: All authors.

    Statistical analysis: Ploubidis.

    Obtained funding: Goodman.

    Conflict of Interest Disclosures: None reported.

    Funding/Support: This study was supported by the Economic and Social Research Council (grants ES/M008584/1 and ES/M001660/1).

    Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

    Additional Contributions: We thank all participating cohort members for their continued commitment to this study.

    References
    1.
    Clark  C, Rodgers  B, Caldwell  T, Power  C, Stansfeld  S.  Childhood and adulthood psychological ill health as predictors of midlife affective and anxiety disorders: the 1958 British Birth Cohort.   Arch Gen Psychiatry. 2007;64(6):668-678. doi:10.1001/archpsyc.64.6.668 PubMedGoogle ScholarCrossref
    2.
    Colman  I, Murray  J, Abbott  RA,  et al.  Outcomes of conduct problems in adolescence: 40 year follow-up of national cohort.   BMJ. 2009;338:a2981. doi:10.1136/bmj.a2981 PubMedGoogle ScholarCrossref
    3.
    Richards  M, Abbott  R. Childhood mental health and adult life chances in post-war Britain: insights from three national birth cohort studies. Smith Institute. Accessed August 24, 2020. http://www.smith-institute.org.uk/wp-content/uploads/2015/10/Childhoodmentalhealthandlifechancesinpostwarbritain.pdf
    4.
    Maughan  B, Stafford  M, Shah  I, Kuh  D.  Adolescent conduct problems and premature mortality: follow-up to age 65 years in a national birth cohort.   Psychol Med. 2014;44(5):1077-1086. doi:10.1017/S0033291713001402 PubMedGoogle ScholarCrossref
    5.
    Jokela  M, Ferrie  J, Kivimäki  M.  Childhood problem behaviors and death by midlife: the British National Child Development Study.   J Am Acad Child Adolesc Psychiatry. 2009;48(1):19-24. doi:10.1097/CHI.0b013e31818b1c76 PubMedGoogle ScholarCrossref
    6.
    Bardone  AM, Moffitt  TE, Caspi  A, Dickson  N, Stanton  WR, Silva  PA.  Adult physical health outcomes of adolescent girls with conduct disorder, depression, and anxiety.   J Am Acad Child Adolesc Psychiatry. 1998;37(6):594-601. doi:10.1097/00004583-199806000-00009 PubMedGoogle ScholarCrossref
    7.
    Winning  A, Glymour  MM, McCormick  MC, Gilsanz  P, Kubzansky  LD.  Psychological distress across the life course and cardiometabolic risk: findings from the 1958 British Birth Cohort Study.   J Am Coll Cardiol. 2015;66(14):1577-1586. doi:10.1016/j.jacc.2015.08.021 PubMedGoogle ScholarCrossref
    8.
    Patalay  P, Moulton  V, Goodman  A, Ploubidis  GB.  Cross-domain symptom development typologies and their antecedents: results from the UK Millennium Cohort Study.   J Am Acad Child Adolesc Psychiatry. 2017;56(9):765-776.e2. doi:10.1016/j.jaac.2017.06.009 PubMedGoogle ScholarCrossref
    9.
    Pinto Pereira  SM, Li  L, Power  C.  Early life factors and adult leisure time physical inactivity stability and change.   Med Sci Sports Exerc. 2015;47(9):1841-1848. doi:10.1249/MSS.0000000000000610 PubMedGoogle ScholarCrossref
    10.
    Maggs  JL, Patrick  ME, Feinstein  L.  Childhood and adolescent predictors of alcohol use and problems in adolescence and adulthood in the National Child Development Study.   Addiction. 2008;103(suppl 1):7-22. doi:10.1111/j.1360-0443.2008.02173.x PubMedGoogle ScholarCrossref
    11.
    Odgers  CL, Moffitt  TE, Broadbent  JM,  et al.  Female and male antisocial trajectories: from childhood origins to adult outcomes.   Dev Psychopathol. 2008;20(2):673-716. doi:10.1017/S0954579408000333 PubMedGoogle ScholarCrossref
    12.
    Goodman  A, Joyce  R, Smith  JP.  The long shadow cast by childhood physical and mental problems on adult life.   Proc Natl Acad Sci U S A. 2011;108(15):6032-6037. doi:10.1073/pnas.1016970108 PubMedGoogle ScholarCrossref
    13.
    Angold  A.  Childhood psychopathology can be really bad for your health.   J Am Acad Child Adolesc Psychiatry. 2009;48(1):3-4. doi:10.1097/CHI.0b013e3181908c49 PubMedGoogle ScholarCrossref
    14.
    Power  C, Elliott  J.  Cohort profile: 1958 British birth cohort (National Child Development Study).   Int J Epidemiol. 2006;35(1):34-41. doi:10.1093/ije/dyi183 PubMedGoogle ScholarCrossref
    15.
    Rutter  M, Tizard  J, Whitmore  K.  Education, Health and Behaviour. Longman Publishing Group; 1970.
    16.
    De Los Reyes  A.  Introduction to the special section: More than measurement error: Discovering meaning behind informant discrepancies in clinical assessments of children and adolescents.   J Clin Child Adolesc Psychol. 2011;40(1):1-9. doi:10.1080/15374416.2011.533405PubMedGoogle ScholarCrossref
    17.
    Van Roy  B, Groholt  B, Heyerdahl  S, Clench-Aas  J.  Understanding discrepancies in parent-child reporting of emotional and behavioural problems: effects of relational and socio-demographic factors.   BMC Psychiatry. 2010;10:56-56. doi:10.1186/1471-244X-10-56 PubMedGoogle ScholarCrossref
    18.
    Yoon  YS, Oh  SW.  Optimal waist circumference cutoff values for the diagnosis of abdominal obesity in Korean adults.   Endocrinol Metab (Seoul). 2014;29(4):418-426. doi:10.3803/EnM.2014.29.4.418 PubMedGoogle ScholarCrossref
    19.
    Lean  MEJ, Han  TS, Morrison  CE.  Waist circumference as a measure for indicating need for weight management.   BMJ. 1995;311(6998):158-161. doi:10.1136/bmj.311.6998.158 PubMedGoogle ScholarCrossref
    20.
    Fuller  E, Power  C, Shepherd  P, Strachan  D. Technical report on the National Child Development Study biomedical survey 2002-2004. National Centre for Social Research. Published July 2006. https://cls.ucl.ac.uk/wp-content/uploads/2017/07/NCDS-biomed-technical-report.pdf
    21.
    Cole  TJ, Altman  DG.  Statistics Notes: Percentage differences, symmetry, and natural logarithms.   BMJ. 2017;358:j3683. doi:10.1136/bmj.j3683PubMedGoogle ScholarCrossref
    22.
    Pinto Pereira  SM, Ki  M, Power  C.  Sedentary behaviour and biomarkers for cardiovascular disease and diabetes in mid-life: the role of television-viewing and sitting at work.   PLoS One. 2012;7(2):e31132. doi:10.1371/journal.pone.0031132 PubMedGoogle Scholar
    23.
    Ki  M, Pouliou  T, Li  L, Power  C.  Physical (in)activity over 20 y in adulthood: associations with adult lipid levels in the 1958 British birth cohort.   Atherosclerosis. 2011;219(1):361-367. doi:10.1016/j.atherosclerosis.2011.07.109 PubMedGoogle ScholarCrossref
    24.
    Geoffroy  MC, Gunnell  D, Power  C.  Prenatal and childhood antecedents of suicide: 50-year follow-up of the 1958 British Birth Cohort study.   Psychol Med. 2014;44(6):1245-1256. doi:10.1017/S003329171300189X PubMedGoogle ScholarCrossref
    25.
    Case  A, Deaton  A.  Rising morbidity and mortality in midlife among white non-Hispanic Americans in the 21st century.   Proc Natl Acad Sci U S A. 2015;112(49):15078-15083. doi:10.1073/pnas.1518393112 PubMedGoogle ScholarCrossref
    26.
    Muthén  B.  A general structural equation model with dichotomous, ordered categorical, and continuous latent variable indicators.   Psychometrika. 1984;49(1):115-132. doi:10.1007/BF02294210 Google ScholarCrossref
    27.
    Rabe-Hesketh  S, Skrondal  A.  Classical latent variable models for medical research.   Stat Methods Med Res. 2008;17(1):5-32. doi:10.1177/0962280207081236 PubMedGoogle ScholarCrossref
    28.
    Oberski  D. Mixture models: Latent profile and latent class analysis. In: Robertson J, Kaptein M, eds.  Modern Statistical Methods for HCI. Springer; 2016:275-287. doi:10.1007/978-3-319-26633-6_12
    29.
    Pang  M, Kaufman  JS, Platt  RW.  Studying noncollapsibility of the odds ratio with marginal structural and logistic regression models.   Stat Methods Med Res. 2016;25(5):1925-1937. doi:10.1177/0962280213505804 PubMedGoogle ScholarCrossref
    30.
    White  IR, Royston  P, Wood  AM.  Multiple imputation using chained equations: issues and guidance for practice.   Stat Med. 2011;30(4):377-399. doi:10.1002/sim.4067 PubMedGoogle ScholarCrossref
    31.
    Frane  AV.  Are per-family type I error rates relevant in social and behavioral science?   J Mod Appl Stat Methods. 2015;14(1):5. doi:10.22237/jmasm/1430453040Google ScholarCrossref
    32.
    Archer  G, Kuh  D, Hotopf  M, Stafford  M, Richards  M.  Adolescent affective symptoms and mortality.   Br J Psychiatry. 2018;213(1):419-424. doi:10.1192/bjp.2018.90PubMedGoogle ScholarCrossref
    33.
    Moffitt  TE, Caspi  A.  Psychiatry’s opportunity to prevent the rising burden of age-related disease.   JAMA Psychiatry. 2019;76(5):461-462. doi:10.1001/jamapsychiatry.2019.0037 PubMedGoogle ScholarCrossref
    34.
    Ploubidis  GB, Sullivan  A, Brown  M, Goodman  A.  Psychological distress in mid-life: evidence from the 1958 and 1970 British birth cohorts.   Psychol Med. 2017;47(2):291-303. doi:10.1017/S0033291716002464 PubMedGoogle ScholarCrossref
    35.
    Collishaw  S, Maughan  B, Natarajan  L, Pickles  A.  Trends in adolescent emotional problems in England: a comparison of two national cohorts twenty years apart.   J Child Psychol Psychiatry. 2010;51(8):885-894. doi:10.1111/j.1469-7610.2010.02252.x PubMedGoogle ScholarCrossref
    36.
    Patalay  P, Gage  SH.  Changes in millennial adolescent mental health and health-related behaviours over 10 years: a population cohort comparison study.   Int J Epidemiol. 2019;48(5):1650-1664. doi:10.1093/ije/dyz006 PubMedGoogle ScholarCrossref
    37.
    Joyce  R, Xu  X. Inequalities in the Twenty-First Century: Introducing the IFS Deaton Review. Institute for Fiscal Studies. Published May 2019. https://www.ifs.org.uk/inequality/wp-content/uploads/2019/05/The-IFS-Deaton-Review-launch.pdf
    38.
    Leon  DA, Jdanov  DA, Shkolnikov  VM.  Trends in life expectancy and age-specific mortality in England and Wales, 1970-2016, in comparison with a set of 22 high-income countries: an analysis of vital statistics data.   Lancet Public Health. 2019;4(11):e575-e582. doi:10.1016/S2468-2667(19)30177-X PubMedGoogle ScholarCrossref
    39.
    Gondek  D, Bann  D, Ning  K, Grundy  E, Ploubidis  GB.  Post-war (1946-2017) population health change in the United Kingdom: A systematic review.   PLoS One. 2019;14(7):e0218991. doi:10.1371/journal.pone.0218991 PubMedGoogle Scholar
    40.
    Birnie  K, Cooper  R, Martin  RM,  et al; HALCyon study team.  Childhood socioeconomic position and objectively measured physical capability levels in adulthood: a systematic review and meta-analysis.   PLoS One. 2011;6(1):e15564. doi:10.1371/journal.pone.0015564 PubMedGoogle Scholar
    41.
    Deary  IJ, Whiteman  MC, Starr  JM, Whalley  LJ, Fox  HC.  The impact of childhood intelligence on later life: following up the Scottish mental surveys of 1932 and 1947.   J Pers Soc Psychol. 2004;86(1):130-147. doi:10.1037/0022-3514.86.1.130 PubMedGoogle ScholarCrossref
    42.
    Lacey  RE, Pinto Pereira  SM, Li  L, Danese  A.  Adverse childhood experiences and adult inflammation: single adversity, cumulative risk and latent class approaches.   Brain Behav Immun. 2020;87:820-830. doi:10.1016/j.bbi.2020.03.017 PubMedGoogle ScholarCrossref
    43.
    Brooks  SK, Webster  RK, Smith  LE,  et al.  The psychological impact of quarantine and how to reduce it: rapid review of the evidence.   Lancet. 2020;395(10227):912-920. doi:10.1016/S0140-6736(20)30460-8 PubMedGoogle ScholarCrossref
    44.
    Galea  S, Merchant  RM, Lurie  N.  The mental health consequences of COVID-19 and physical distancing: the need for prevention and early intervention.   JAMA Intern Med. 2020. doi:10.1001/jamainternmed.2020.1562 PubMedGoogle Scholar
    45.
    Kendler  KS, Gardner  CO.  Sex differences in the pathways to major depression: a study of opposite-sex twin pairs.   Am J Psychiatry. 2014;171(4):426-435. doi:10.1176/appi.ajp.2013.13101375 PubMedGoogle ScholarCrossref
    46.
    Pongiglione  B, De Stavola  BL, Kuper  H, Ploubidis  GB.  Disability and all-cause mortality in the older population: evidence from the English Longitudinal Study of Ageing.   Eur J Epidemiol. 2016;31(8):735-746. doi:10.1007/s10654-016-0160-8 PubMedGoogle ScholarCrossref
    47.
    Verbrugge  LM.  Sex differentials in health.   Public Health Rep. 1982;97(5):417-437.PubMedGoogle Scholar
    48.
    Patton  GC, Coffey  C, Romaniuk  H,  et al.  The prognosis of common mental disorders in adolescents: a 14-year prospective cohort study.   Lancet. 2014;383(9926):1404-1411. doi:10.1016/S0140-6736(13)62116-9 PubMedGoogle ScholarCrossref
    49.
    VanderWeele  TJ, Ding  P.  Sensitivity analysis in observational research: introducing the E-value.   Ann Intern Med. 2017;167(4):268-274. doi:10.7326/M16-2607 PubMedGoogle ScholarCrossref
    50.
    Ding  P, VanderWeele  TJ.  Sensitivity analysis without assumptions.   Epidemiology. 2016;27(3):368-377. doi:10.1097/EDE.0000000000000457 PubMedGoogle ScholarCrossref
    51.
    Rose  G.  Sick individuals and sick populations.   Int J Epidemiol. 1985;14(1):32-38. doi:10.1093/ije/14.1.32 PubMedGoogle ScholarCrossref
    ×