Twenty-Five and Up (25Up) Study: A New Wave of the Brisbane Longitudinal Twin Study

The aim of the 25 and Up (25Up) study was to assess a wide range of psychological and behavioral risk factors behind mental illness in a large cohort of Australian twins and their non-twin siblings. Participants had already been studied longitudinally from the age of 12 and most recently in the 19Up study (mean age = 26.1 years, SD = 4.1, range = 20–39). This subsequent wave follows up these twins several years later in life (mean age = 29.7 years, SD = 2.2, range = 22–44). The resulting data set enables additional detailed investigations of genetic pathways underlying psychiatric illnesses in the Brisbane Longitudinal Twin Study (BLTS). Data were collected between 2016 and 2018 from 2540 twins and their non-twin siblings (59% female, including 341 monozygotic complete twin-pairs, 415 dizygotic complete pairs and 1028 non-twin siblings and singletons). Participants were from South-East Queensland, Australia, and the sample was of predominantly European ancestry. The 25Up study collected information on 20 different mental disorders, including depression, anxiety, substance use, psychosis, bipolar and attention-deficit hyper-activity disorder, as well as general demographic information such as occupation, education level, number of children, self-perceived IQ and household environment. In this article, we describe the prevalence, comorbidities and age of onset for all 20 examined disorders. The 25Up study also assessed general and physical health, including physical activity, sleep patterns, eating behaviors, baldness, acne, migraines and allergies, as well as psychosocial items such as suicidality, perceived stress, loneliness, aggression, sleep–wake cycle, sexual identity and preferences, technology and internet use, traumatic life events, gambling and cyberbullying. In addition, 25Up assessed female health traits such as morning sickness, breastfeeding and endometriosis. Furthermore, given that the 25Up study is an extension of previous BLTS studies, 86% of participants have already been genotyped. This rich resource will enable the assessment of epidemiological risk factors, as well as the heritability and genetic correlations of mental conditions.

Mental health; cohort study; longitudinal; genetics (Received 20 April 2019; accepted 13 May 2019) The Twenty-Five Up Study (25Up: Can we predict who will develop mental disorders? A long-term study of adolescent twins) ran from 2016 to 2018 and assessed a wide range of mental health and behav- ioral disorders in a young Australian cohort consisting of twins and their non-twin siblings. This study is an extension of its predecessor, the 19Up study, and the Brisbane Longitudinal Twin Study (BLTS; Couvy-Duchesne et al.,2018;G i l l e s p i ee ta l . ,2013), which examined mental health and alcohol and substance use (Gillespie et al., 2009) in a large sample of young twin adults (mean age = 26.1, range = 19.7–38.6). 25Up aimed to assess individuals who were slightly older than those in 19Up, with the similar overall aim of collecting data that will help shed light on the risk factors and pathways involved in the development of affective disorders. These studies re- present a large collection of a wide array of phenotypic data, includ- ing psychological and environmental variables of relevance for mental health and disorders. Some of these include personality dimensions and psychological symptoms, mental disorders that meet the Diagnostic and Statistical Manual of Mental Disorders (5th ed., DSM-5; American Psychiatric Association, 2013) and five diagnostic criteria: alcohol and substance use and misuse, migraine, sleep behaviour, as well as neurobiological correlates (neuroimag- ing) and genome-wide genotyping ( Figure 1 ). Notably, the mean age increase between the individuals participating in the 19Up and 25Up studies was 3.6 years (Supplementary Figure 1). Author for correspondence: Brittany L. Mitchell, Email: Brittany.mitchell@ qimrberghofer.edu.au *Authors contributed equally. Cite this article: Mitchell BL, Campos AI, Rentería ME, Parker R, Sullivan L, McAloney K, Couvy-Duchesne B, Medland SE, Gillespie NA, Scott J, Zietsch BP, Lind PA, Martin NG, and Hickie IB. (2019) Twenty-Five and Up (25Up) Study: A New Wave of the Brisbane Longitudinal Twin Study. Twin Research and Human Genetics 22: 154 –163, https://doi.org/10.1017/thg.2019.27 © The Author(s) 2019. Twin Research and Human Genetics (2019), 22, 154 –163 doi:10.1017/thg.2019.27 https://doi.org/10.1017/thg.2019.27 Published online by Cambridge University Press The 25Up study was structured so as to collect further psycho- logical, behavioral and subclinical assessments, with the aim of using these to understand the factors involved in the development and progression of mental health disorders from late teenage years to early adulthood. This new wave of the BLTS will allow us to gain insight into the genetic basis of behavioral phenotypes, mental health and comorbidities. Finally, as with the 19Up study, data col- lection was also designed to contribute to twin and genetic consor- tia in psychiatry, personality and behavior genetics. The aim of the present article is to describe the key characteristics of the cohort and give examples of important themes (e.g. comorbidity and pri- macy of onset in dual diagnoses) that can be examined in future in more detailed analyses of the cohort data, and also to highlight the opportunities for new collaborations that could make use of the rich 25Up data set.

Study Contact Between February 2016 and October 2018, 3785 individuals were approached via email to participate in the 25Up study, which entailed a detailed, three-part, self-report online survey (TFU1, TFU2, and TFU3). There was a 67.1% response rate (2540 individ- uals) for completing the first part of the survey (TFU1) and a 62% response rate (2343 individuals) for completing all three parts of the survey (TFU1, TFU2, and TU3). Follow-up was conducted by a phone call 1 –2 weeks after the initial contact email. Supplementary Figure 2 depicts the variation in participant response. Part 1 of the survey (TFU1, N = 2540) assessed general health, medical and treatment history as well as lifetime diagnoses. TFU2 (N = 2484) expanded on TFU1, assessing other physical character- istics such as baldness, skin tone and acne, as well as a multitude of conditions and behavioral characteristics. TFU3 ( N = 2343) was the final part of the survey, containing questions addressing sexual identity, quality of romantic relationships, sexual and romantic preferences. For further information regarding the specific items addressed in each section, the instruments used to assess these items and number of participants answering each section, refer to Table 1 . Statistical Analyses In this study, we focus on describing the prevalence, risk factors (i.e. age and sex), comorbidity and primacy of onset of self- reported mental health variables. A study assessing specific diag- noses using the relevant questionnaires is outside the scope of this cohort description. Age and sex effects were assessed by means of logistic regression modeling in Python using the statsmodels mod- ule. The variables of interest were included as covariates and their significance was assessed using Wald tests. Disorder Clustering and Comorbidity Analysis All pairwise tetrachoric correlations ( ρ) were calculated using R (3.1.1) and the psych package (v1.4.3). Briefly, the presence or absence of a disorder was coded as a binary vector, and all disorders were compared by computing their tetrachoric correlation coeffi- cient. Next, a hierarchical clustering was performed using Ward ’s minimum intracluster variance objective function (Ward, 1963) with a distance metric calculated as 1 − ρ. Mixed-effects logistic regressions were used to calculate the increase in risk on a second disorder given the occurrence of a given disorder. The function glmer from the R library lme was used, including age and sex as fixed effects, and the family ID as grouping factor for the random effects variance –covariance structure. For each pair of disorders, one was modeled as an outcome variable, while the other one was used as a predictor. The p-values of the association of one dis- order with another were subjected to a Bonferroni multiple testing correction. Fig. 1. Summary of the BLTS data collection. Longitudinal: vitamin D; infections (anti- bodies); neuroticism junior Eysenck personality questionnaire (JEPQ) Neuroticism – Extraversion–Openness inventory (NEO); psychiatric signs (SPHERE). Cross-sectional: hair cortisol; cognition (verbal, performance IQ, working memory, and information processing); binocular rivalry (rivalry rate); brain imaging (multimodal magnetic res- onance imaging); substance use (alcohol, tobacco, and recreational drugs); sleep pat- terns (actigraphy); psychiatric diagnoses (Composite International Diagnostic Interview); life events/social support/relationships (e.g. early home environment, fam- ily relationships, traumatic events, socioeconomic factors). Note: Sample size is only indicative as some of the waves are still recruiting new participants. Figure adapted from Couvy-Duchesne et al. ( 2018) and Gillespie et al. ( 2013). Twin Research and Human Genetics 155 https://doi.org/10.1017/thg.2019.27 Published online by Cambridge University Press Fig. 2. Kaplan–Meier curves stratified by sex. Kaplan–Meier curves depict the self-reported age of onset for three disorders (depression, anorexia and bipolar disorder, respectively) in the 25Up cohort. Table 1. Variables examined in the different surveys of the 25Up study TFU1 Instrument Approximate N Demographics Young and Well Cooperative Research Centre (YAW CRC) 2540 General health and well-being QIMR 16Up 682 –2535 Puberty Perceived pubertal timing (Cance et al., 2012) 1310–2476 Cognition Self-perceived 2481 Medical history National Comorbidity Survey (NCS) screener 1949 Morning sickness Nausea and Vomiting of Pregnancy (NVP) Short form 1777 Breast feeding NVP short form 550 Pregnancy-related depression screen 300 –655 Mental health Global Assessment of Functioning (GAF) 2489 Mental health NCS Screener 2489 Mental health Treatment history 874 Alcohol/substances Treatment history 843 Family mental health YAW CRC 2379 Lifetime experiences — anxiety disorders NCS screener 2455 Panic disorder NCS screener 2493 Social phobia NCS screener 1563 Specific phobia NCS screener 2493 Separation anxiety disorder NCS screener 2370 Major Depressive Disorder (MDD) NCS screener 2500 Mania–hypomania NCS screener 2445 Irritable depression NCS screener 2488 Oppositional defiance disorder NCS screener 2487 Conduct disorder NCS screener 2513 ADHD NCS screener 2298 Intermitent explosive disorder NCS screener 2512 (Continued) Table 1. (Continued ) TFU1 Instrument Approximate N Current mental health BMRI 2536 Current mental health Kessler 10 2538 Day functioning Composite International Diagnostic Interview (CIDI) 2529 Personality NEO TIPI Scale 2536 Hypo mania screener Five-item self-report 2402 Psychosis screener Self-rated scale based on CIDI 2513 Functional impairment GAF 1442 Functional impairment Social and Occupational Functioning Assessment Scale 1442 Suicidality Self-harm/past month/lifetime 2531 Alcohol or other substance misuse From QIMR 16Up 2369 Physical health and activity schedule Australian Bureau of Statistics National Survey 2200 Sleep–wake cycle BMRI from 19Up 2489 Eating behaviors YAW CRC 2465 Body image YAW CRC 2534 Relationships and social networking QIMR 16Up 2514 TFU2 Instrument Approximate N Demographics being a twin NA 2141 Physical phenotypes — Part 1 HWB 1490 Physical phenotypes — Part 2 HWB 2468 Migraines HWB 2434 Migraines — females HWB 566 Asthma eczema allergies HWB 2403 Endometriosis HWB 1497 (Continued) 156 Brittany L. Mitchell et al. https://doi.org/10.1017/thg.2019.27 Published online by Cambridge University Press Power Analysis Phenotype simulations (of continuous traits) and power analyses were performed using the powerFun, MASS and OpenMx R libra- ries. Multivariate data were simulated using the mvrnorm function specifying a variance –covariance structure determined by a linear combination of varying values for the additive genetic (A) and common environmental (C) components. The sample sizes of these simulated distributions were identical to the number of twin-pairs available in the 25Up cohort. OpenMx was used to fit an ACE model to the simulated data. The significance of the A and C variance components was assessed using a log likelihood ratio test (the mxCompare function of OpenMx) comparing the full model to a model only including the other two components (e.g. ACE vs. CE). This procedure was repeated 100 times, and power was estimated by counting the number of iterations in which the studied component was rejected. A similar procedure was used to simulate and assess the power to detect a genetic correlation. Two phenotypes were simulated with a specified underlying genetic correlation, and a bivariate ACE model was used to assess the significance of the genetic correlation. The results are available in the Supplementary Figure 3 .

Cohort Description Of the 3785 individuals invited to participate in the study, 62% of the twins and non-twin siblings provided complete data. Overall, females were slightly over-represented among the 25Up respon- dents, comprising 52% of the invited population but 59.5% of actual ascertained participants ( Table 2). Survey completion rates were high and, of the participants who had completed TFU1, 2484 (97.8%) completed the second section of the survey (TFU2), and 2343 (92.2%) completed the third part (TFU3). Females tended to complete all sections of the survey more often than males, with 95.3% of females completing all three parts compared to 87.7% of males. The greatest dropout for men was between completing Part 2 (96.3%) and Part 3 (87.7%) of the survey. The mean age of all participants was 29.7 (SD = 4.2, range = 22–44; Supplementary Figure 2), consisting of 341 complete monozygotic pairs, 415 dizy- gotic pairs, 125 MZ singletons, 269 DZ singletons and 634 siblings. Twins and non-twin siblings did not differ in maximum educa- tional attainment level ( p = .57), but nontwin individuals were older (30.5 vs. 29.4, p = .001), more likely to be married (62% vs. 55%, p = .001) and less likely to have children compared with co-twins (49.6% vs. 57.5%, p = .001). Ethnically, the cohort reflects the population structure of families with twins in Queensland at the time of recruitment, with a majority of participants having European ancestry and minorities of predominantly Asian ances- try (Gillespie et al., 2013). All participants had been invited to complete previous BLTS (Gillespie et al., 2013; Wright & Martin, 2004) studies ( Figure 1). Therefore, variables such as height, weight, personality, psychiatric signs, sleep patterns, migraine and blood samples (hematological and immunological measures: e.g. antibodies markers of infections, vitamin D) were collected longitudinally in the BLTS, with up to five time points for some phenotypes (Figure 1). A noteworthy example is the assessment of personality traits using the Neuroticism – Extraversion–Openness (NEO) Personality Inventory-related scales (Costa & McRae, 1992). Although some cohorts present different versions of the NEO (due to updates and study design changes), the overall constructs measured should remain highly isomorphic, Table 1. (Continued ) TFU1 Instrument Approximate N Physical Phenotypes — Part 3 HWB 2449 Current mental health Perceived stress scale 2464 Current mental health Borderline, autism and loneliness (PAI/BOR and SRS) 2460 Current mental health Adult ADHD Self-Report 2440 Current mental health Buss Perry Aggression Questionnaire 2432 Sleep–wake cycle Pittsburgh Sleep Quality Assessment) 2424 Sleep apnea screen Maislin et al., 1995 2348 Caffeine and general sleep questions NA 2413 Sleep–wake cycle Insomnia Severity Index 2379 Eating behavior and anorexia nervosa From QIMR 16Up 2394 Social networking and relationships PBI 2321 Social networking and relationships Kessler perceived social support 2241 Stressful life events List of threatening experiences 2301 Technology use YAW CRC 2405 Games and gambling Problem gambling severity index 1942 Cyberbullying and sexting NA 2160 TFU3 Instrument Approximate N Demographics NA 2331 Romantic preferences Designed by Zietsch (from 19Up) 2324 Romantic preferences Fluid gender identity based on Multi-GIQ (Joel et al., 2014) 2295 Romantic preferences — females Contraceptives 1430 SNR-disgust Three domain disgust scale 2280 Sociosexuality NA 2290 Self-rated physical attractiveness NA 2286 Attraction NA 2254 Relationships NA 1736 Partner section Cognition and self-report IQ, education level, SPHERE, height and weight and eye color 9 Note: TFU1, TFU2 and TFU3 refer to the three parts of the online questionnaire. Approximate N represents the average of not null respondents for representative (not follow-up) questions of each section. Twin Research and Human Genetics 157 https://doi.org/10.1017/thg.2019.27 Published online by Cambridge University Press Fig. 3. Disorder comorbidity within the 25Up study. Lower triangle depicts a hierarchical clustering (Ward ’s method) of the disorders based on their self-reported lifetime co- occurrence (tetrachoric correlations). Upper triangle por- trays lifetime comorbidity odds ratio (ordered based on the clustering of the lower triangle). Note: *p < .05 after multiple testing correction ( α < .000146). Table 2. Demographics of the 25Up cohort Total Females Males Completed Part 1 (TFU1) 2540 (100.0%) 1523 (100.0%) 1017 (100.0%) Completed Part 2 (TFU2) 2484 (97.8%) 1505 (98.8%) 979 (96.3%) Completed Part 3 (TFU3) 2343 (92.2%) 1451 (95.3%) 892 (87.7%) Age (SD; range) 29.7 (4.2; 22 –44] 29.4 (4.3; 22 –44) 29.9 (4.2; 22 –41) Single 679 (26.7%) 393 (25.8%) 286 (28.1%) Married 1502 (59.1%) 916 (60.1%) 586 (57.6%) Relationship 296 (11.7%) 175 (11.5%) 121 (11.9%) Separated but married 35 (1.4%) 22 (1.4%) 13 (1.3%) Divorced 23 (0.9%) 12 (0.8%) 11 (1.1%) Widowed 0 (0.0%) 0 (0.0%) 0 (0.0%) No formal education 10 (0.4%) 6 (0.4%) 4 (0.4%) Primary school 0 (0.0%) 0 (0.0%) 0 (0.0%) Junior high school 36 (1.4%) 10 (0.7%) 26 (2.6%) Senior high school 259 (10.2%) 156 (10.2%) 103 (10.1%) Certificate or diploma 679 (26.7%) 358 (23.5%) 321 (31.6%) Degree 1,054 (41.5%) 680 (44.6%) 374 (36.8%) Postgraduate diploma, masters or PhD 500 (19.7%) 312 (20.5%) 188 (18.5%) 158 Brittany L. Mitchell et al. https://doi.org/10.1017/thg.2019.27 Published online by Cambridge University Press as we would expect with biometrical phenotypes (such as height) and other behavioral instruments such as the Somatic and Psychological HEalth REport (SPHERE; Hickie et al., 2001)( a l s o used to assess mental health on most of the BLTS). In addition, genome-wide single nucleotide polymorphism genotypes are cur- rently available for 86% ( N = 2205) of participants. Findings to Date The 25Up study has collected information on 20 different psychi- atric or affective disorders (see Table 3 ) and a range of lifestyle, health and behavioural traits ( Table 1). Overall, ∼20% of the par- ticipants self-reported a lifetime major mental health problem affecting their everyday life. This estimate is consistent with esti- mates for the Australian population (Department of Health, 2009). Among the disorders examined, general anxiety (16%, N = 402) and depression (17.3%, N = 436) were the most prevalent diagnoses. Following these, panic, substance use, sleep, and post traumatic stress disorders (PTSD) had the highest prevalence in the 25Up cohort (3.3%, 2.9%, 2.5% and 2.4%, respectively;Table 3). Sex Differences The differences in response rates and fallout rates between males and females motivated the assessment of whether self-reported life- time prevalence of mental health problems was associated with sex in this cohort. Both depression and general anxiety were far more prevalent in females than males (11.6% vs. 5%, p < .001 and 15.3% vs. 3.4%, p < .001, respectively). PTSD, obsessive-compulsive dis- order (OCD), panic disorder, general eating disorders and bulimia and anorexia showed significant sex effects, all having a higher prevalence in females ( Table 3 ). In addition, PTSD was the only disorder to show a nominally significant increased prevalence with age (p = .006), although this did not survive correcting for multiple testing, but would be consistent with a higher probability for the occurrence of a traumatic event as time passes. No other differences reached statistical significance in this cohort ( Table 3). Age of Onset The age of onset of the examined self-reported phenotypes (disor- ders) was not significantly different between males and females. In the case of depression, females had a slightly earlier age of onset (18.8 years vs. 20.5 years, p ≤ .001; Figure 2 and Supplementary Figure 4 ). The youngest mean age of onset was for autistic spec- trum disorders, including Asperger syndrome (mean = 5.5 years, SD = 6.5), while the oldest mean age of onset recorded in this cohort was for psychosis (mean = 24.4 years, SD = 5.9). Age of onset estimates was not available for alcohol dependence and mis- use as only the age at alcohol drinking initiation (mean = 16.0 years) was collected. Notably, the mean age of onset of anorexia, bulimia and eating disorders were all during adolescence ( ∼16 years), while the mean age of onset for other disorders was mostly around young adulthood ( Table 4 and Supplementary Figure 4 ). Disorder Comorbidity There is a known overlap between affective, anxiety and substance use disorders (Kessler et al., 1996; Merikangas et al., 1996; Regier et al., 1990). Within the 25Up study, evidence of the relationships between the 20 disorders was observed through hierarchical clus- tering. Self-reported history of psychosis and schizophrenia Table 3. Lifetime prevalence of self-reported mental health disorders in the 25Up study N Prevalence (%)* Male Prevalence (%) Female Prevalence (%) OR —95%C.I.— Sex pvalue Age pvalue AnyMHP 519 20.6 18.6 21.7 0.030 0.002 ADD ADHD 32 1.3 1.5 1.1 0.360 0.060 Anorexia 37 1.5 0.3 2.2 0.001 0.190 Anxiety 402 16.0 8.9 20.5 0.000 0.140 Au/g415sm 13 0.5 0.9 0.3 0.040 0.410 Bipolar Disorder 28 1.1 1.2 1.1 0.740 0.730 Bulimia 27 1.1 0.1 1.7 0.005 0.310 Conduct Disorder 2 0.1 0.1 0.1 0.750 0.690 Depression 436 17.3 12.9 20.0 0.000 0.290 Drug addic/g415on 38 1.5 1.3 1.6 0.460 0.840 Ea/g415ng disorder 57 2.3 0.1 3.7 0.000 0.210 Memory disorder 12 0.5 0.6 0.4 0.500 0.650 Narcolepsy 1 0.0 0.1 0.0 1.000 0.380 OCD 45 1.8 0.7 2.5 0.002 0.510 PMS 28 1.1 0.0 1.8 1.000 0.460 PTSD 61 2.4 1.4 3.1 0.005 0.004 Panic 83 3.3 1.5 4.5 0.000 0.330 Psychosis 12 0.5 0.5 0.5 0.920 0.830 SUD 73 2.9 3.0 2.8 0.910 0.190 Schizophrenia 9 0.4 0.4 0.3 0.790 0.880 Sleep disorder 63 2.5 2.2 2.7 0.290 0.290 Nominally significant P values are highlighted in bold. Analyses were performed by using a logis/g415c regression accoun/g415ng simultaneously for the effects of sex (females as a reference) and age. OR - odds ra/g415o, C.I.- 95% confidence intervals. MHP- mental health problem. ADD/ADHD- A/g425en/g415on deficit disorder/A/g425en/g415on deficit and hyperca/g415vity disorder. OCD- obsessive compulsive disorder. PMS - premenstrual syndrome. PTSD - post trauma/g415c stress disorder. SUD- Alcohol and substance misuse. , MHP- mental health problem. *Prevalences calculated based only on not null values (par/g415cipants that responded to the sec/g415on (N=2516) Twin Research and Human Genetics 159 https://doi.org/10.1017/thg.2019.27 Published online by Cambridge University Press clustered together, as did anorexia, bulimia and eating disorders. Additional clusters were attention deficit and hypercativity disor- der (ADHD) and autism, as well as depression, anxiety, PTSD and panic disorder (Figure 3). In order to quantify the increased risk of a condition, given the presence of a second condition (while cor- recting for the effects of age, sex and relatedness), we used a mixed- effects logistic regression approach (see Methods). All significant associations between the disorders studied were positive while none of the negative associations (i.e. decreased risk) reached stat- istical significance. Individuals who were rated as positive for an alcohol and other substance misuse were more likely to score pos- itive for memory disorder and depression ( p < .05 after multiple testing correction). Furthermore, a significant risk increase was also detected between psychosis and schizophrenia and between depression and a variety of other comorbid disorders, including PTSD, sleep disorder and premenstrual syndrome (PMS) among others ( Figure 3 ). Interestingly, of those participants (both males and females) reporting substance misuse ~50% reported also hav- ing another mental health disorder, the majority reported that the substance abuse disorder followed the mental condition (Figure 4). The 25Up Cohort Will Enable Longitudinal Analyses The unique strength of the 25Up study is that it is the latest wave in a longitudinal study spanning more than 20 years. This allows for unparalleled analysis of the dynamic nature of mental health var- iables as individuals progress through adolescence and into young adulthood. For example, when comparing the lifetime prevalence of self-reported psychotic symptoms (CIDI Psychosis Screener; Scott et al., 2006) in the previous 19Up cohort to those in 25Up, we found that, as expected, the prevalence for most symptoms has increased in the 25Up. However, there were instances where the lifetime prevalence decreased in the 25Up cohort, pointing to possible recall bias. Nonetheless, with those that increased, the increase was heterogeneous, that is, the prevalence of some symptoms did not change significantly, whereas others doubled (Figure 5 ). The extent to which this heterogeneity is caused by recall bias or other factors might be studied in the future. Notably, the BLTS includes several potential isomorphic instru- ments (such as depression and personality) that will enable genetic and environmental longitudinal analyses.

Here we described the demographics and self-reported history of mental disorders in the 25Up cohort. Our findings are consistent with previous mental illness prevalence estimates in Australia (Lawrence et al., 2016; Liddell et al., 2016), and observations of women having a higher prevalence of PTSD (Galea et al., 2005; Gavranidou & Rosner, 2003), panic disorder (Crowe et al., 1983; Weissman et al., 1997) eating disorders (Mitchison & Hay, 2014), depression (Weissman & Klerman, 1977) and anxiety disorders (Bandelow & Michaelis,2015). We detected a significant association between sex and OCD prevalence, an observation not previously made in adults (Karno et al., 1988;L ´opez-Solà et al., 2014), but only in individuals with an age of onset before or during adolescence (Grant, 2014). We also identified a nominal association of PTSD with age, although a plausible explanation of this effect could be related to a higher probability of a stressful event with age, but this association did not survive multiple testing corrections. There is consistent evidence for a broad distinction between externalizing and internalizing disorders (Cosgrove et al., 2011; Krueger et al., 1998). Nonetheless, our findings suggest high levels of lifetime comorbidity between affective and substance use disor- ders. This is consistent with recent studies detecting a genetic over- lap between psychological distress, somatic distress, affective disorders and substance use (Chang et al., 2018) and with the self-medication hypothesis suggesting substance misuse as a coping mechanism (Marshall, 1994; Myrick & Brady, 2003). Notably, we identified a high comorbidity between externalizing Fig. 4. Most participants reported a mental disorder prior to substance abuse. Bar plots depicting the number of participants reporting precedence of either a mental or a substance abuse disorder. Only participants that reported both type of conditions responded to this question ( n = 66). Table 4. Disorder age of onset in the 25Up study Trait Males Females p-value ADD/ADHD 9.91 (3.2) 10.4 (6.09) 0.8 Anorexia 23.5 (6.5) 16.6 (5.1) 0.1 Anxiety 19.2 (6.7) 18.4 (6.4) 0.4 Autism/Asperger 4.0 (4.53) 8.5 (8.5) 0.5 Bipolar disorder 21.6 (5.5) 19.5 (4.3) 0.3 Bulimia 30.0 (0.0) 16.4 (2.2) NA Conduct disorder 6.0 (0.0) 11.0 (0.0) NA Depression 20.5 (6.6) 18.8 (5.6) 0.01 Drug addiction 18.1 (2.8) 18.5 (3.3) 0.7 Eating disorder 29.0 (0.0) 16.5 (4.1) NAN Memory disorder 19.5 (4.0) 22.8 (8.8) 0.6 Narcolepsy NA NA NA OCD 20.7 (4.5) 16.8 (6.6) 0.2 PMS NA 17.5 (5.8) NA PTSD 22.8 (5.4) 22.5 (6.5) 0.9 Panic 22.2 (6.2) 19.7 (7.1) 0.2 Psychosis 23.8 (3.1) 24.8 (7.4) 0.8 Schizophrenia 19.7 (2.4) 20.8 (6.6) 0.8 Sleep disorder 21.8 (6.0) 20.9 (8.0) 0.7 Alcohol and substance misuse 20.0 (4.6) 18.6 (4.1) 0.2 Note: ADD/ADHD = attention deficit disorder/attention-deficit hyper-activity disorder; OCD = obsessive-compulsive disorder; PMS = premenstrual syndrome; PTSD = post traumatic stress disorder. 160 Brittany L. Mitchell et al. https://doi.org/10.1017/thg.2019.27 Published online by Cambridge University Press and internalizing disorders such as depression and substance misuse or OCD and attention deficit disorder (ADD)/ADHD. Furthermore, we detected expected comorbidity of co-occurring diseases and symptoms such as eating disorders and anorexia (Thornton et al., 2010), drug addiction and alcohol misuse, psy- chosis and schizophrenia (National Collaborating Centre for Mental Health 2014) and depression and anxiety (Gorman, 1996). Depression was the disorder with the highest comorbidity, having a significant association with around half of the conditions. Altogether, these observations are consistent with the high genetic correlation between psychiatric disorders and the overlap in diag- nostic criteria (Anttila et al., 2018; Bulik-Sullivan et al., 2015). Furthermore, these findings suggest the self-reported lifetime of mental health conditions on the 25Up to be valid and reliable. We also analyzed whether there were sex differences in the age of onset of psychiatric and substance abuse disorders. Our findings indicated a nominally significant difference in the age of onset of depression, which is consistent with previous observations on ado- lescents (Avenevoli et al., 2015). Notably, the mean age of onset of eating disorders was during adolescence, as previously reported (Hudson et al., 2007; Volpe et al., 2016). Strengths and Limitations The 25Up study represents a large effort to characterize and under- stand the genetic, environmental and behavioral factors associated with mental health in young adults transitioning from adolescence. More than 50 different mental health and lifestyle variables, such as technology use, sociosexuality and substance use, were assessed in (on average) ~2100 twins, making this a rich and valuable data set. The preceding work on the BLTS cohort has started to shed light onto the genetic and neurobiological etiology of substance abuse (Chang et al., 2018; Gillespie et al., 2018; Schmaal et al., 2016). We expect that this follow-up of the cohort will drive further cross-sectional and longitudinal genetic epidemiology studies of human behaviour and mental health. The sample size of the 25Up study (341 complete monozygotic twin-pairs and 415 dizygotic pairs) should allow the detection of heritability estimates ≥30% and common environmental influences >30% with at least 80% power (Martin et al., 1978; Visscher et al., 2008). Furthermore, the 25Up cohort provides at least 80% power to detect genetic correlations as low as r g = .3 (considering a heritability for each trait >20%) in line with cohorts with similar sizes and power simulations. The existence of longi- tudinal biometric and mental health data could enable novel family and epidemiological studies to be performed. The longitudinal nature of the BLTS has enabled the discovery that lifetime progres- sion of self-reported psychotic symptoms is heterogeneous. Future studies assessing the genetic and environmental factors accounting for this heterogeneity across development could result in new intervention and prevention strategies. Some limitations of the 25Up study must be noted. Although clinical, lifestyle and demographic variables were assessed through established instruments, they were obtained through online sur- veys and therefore all responses are subject to the possible biases and accuracy of self-report questionnaires. Notably, recall bias (e.g. not remembering a depressive episode) and subjectivity (e.g. when rating their physical and mental health) should be con- sidered and corrected for, before conducting analyses and espe- cially before comparing with other population-based studies. It is also important to note that all analyses presented in this article were based on self-reported lifetime medical history data and therefore are not necessarily in accordance with observations made when using other criteria, such as theDSM-5. Additionally, caution must be taken when trying to ascertain specific disorders from the instruments used in this study. For example, with schizophrenia, self-reported and CIDI-based ascertainment might not match the DSM-5 diagnostic criteria. Fig. 5. Comparison of lifetime prevalence of any psychotic symptoms in the 19Up and 25Up cohort studies. Bar plots depict the prevalence and 95% confidence intervals (1000 bootstrap pseudo replications) of psychotic symptoms in the 19Up and 25Up cohorts stratified by sex. Results are depicted only for participants with available data for both data sets ( n = 2319; M = males, F = females). Twin Research and Human Genetics 161 https://doi.org/10.1017/thg.2019.27 Published online by Cambridge University Press Furthermore, genetic and epidemiological analyses using the twins from the 25Up cohort will inevitably assume the cohort to be representative of the overall (age-matched) population. It is well documented that the liability to twinning has a genetic component (Mbarek et al., 2016; Painter et al., 2010) and could therefore be genetically correlated with other traits (Laisk et al., 2018). Moreover, environmental differences between twins and nontwins (such as twins being treated more similar than siblings) are also a possible source of bias. Nonetheless, both of these limitations can be circumvented by including the large number of siblings who are part of the 25Up cohort in the analyses.

The 25Up cohort represents a rich data set that will enable analyses of the epidemiology, heritability and genetic correlations of well- being and mental health variables. The overall prevalence of mental disorders and the prevalence rates for lifetime comorbidities were in line with previous studies. As an update of the BLTS, it represents a unique opportunity for longitudinal studies aiming at further under- standing the etiology and heterogeneity underlying the progression of affective, mood and substance use disorders from young- to mid- adulthood. Here, we exemplified this by comparing item-level prevalence of psychotic items across the 19Up and 25Up cohorts, identifying a heterogeneous increase that might be explained by dynamic (age-dependent) genetic effects. Moreover, the inclusion of surveys focused on the usage of technological devices (e.g. internet and mobile phone usage) will enable unprecedented analysis of their relationship with mental health. We anticipate the 25Up cohort to be an attractive resource to boost collaboration, ultimately propelling scientific discovery. Supplementary Material. To view supplementary material for this article, please visit https://doi.org/10.1017/thg.2019.27. Acknowledgments. This project was funded by NHMRC Project Grant APP1069141 to IBH and NGM. We also thank David Smyth and Scott Gordon for IT support. In particular, we thank the twins and their families for their participation in our research. AIC is supported by a research training scholarship (RTP) granted by The University of Queensland (UQ).

American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.). Washington, DC. Anttila, V., Bulik-Sullivan, B., Finucane, H. K., Walters, R. K., Bras, J., Duncan, L., ::: Malik, R. (2018). Analysis of shared heritability in common disorders of the brain. Science, 360, eaap8757. Avenevoli, S., Swendsen, J., He, J.-P., Burstein, M., & Merikangas, K. R. (2015). Major depression in the National Comorbidity Survey — Adolescent Supplement: Prevalence, correlates, and treatment. Journal of the American Academy of Child & Adolescent Psychiatry , 54,3 7–44. Bandelow, B., & Michaelis, S. (2015). Epidemiology of anxiety disorders in the 21st century. Dialogues in Clinical Neuroscience , 17, 327 –335. Bulik-Sullivan, B., Finucane, H. K., Anttila, V., Gusev, A., Day, F. R., Loh, P.-R., ::: Robinson, E. B. (2015). An atlas of genetic correlations across human diseases and traits. Nature Genetics, 47, 1236. Cance, J. D., Ennett, S. T., Morgan-Lopez, A. A., & Foshee, V. A. (2012). The stability of perceived pubertal timing across adolescence. Journal of Youth and Adolescence, 41, 764–775. Chang, L.-H., Couvy-Duchesne, B., Medland, S. E., Gillespie, N. A., Hickie, I. B., Parker, R., & Martin, N. G. (2018). The genetic relationship between psychological distress, somatic distress, affective disorders, and substance use in young australian adults: A multivariate twin study. Twin Research and Human Genetics, 21, 347 –360. Cosgrove, V. E., Rhee, S. H., Gelhorn, H. L., Boeldt, D., Corley, R. C., Ehringer, M. A., ::: Hewitt, J. K. (2011). Structure and etiology of co-occurring internalizing and externalizing disorders in adolescents. Journal of Abnormal Child Psychology , 39, 109–123. Costa, P. T., & McCrae, R. R. (1992). Normal personality assessment in clinical practice: The NEO Personality Inventory. Psychological Assessment, 4,5 –13. Couvy-Duchesne, B., O ’Callaghan, V., Parker, R., Mills, N., Kirk, K. M., Scott, J., ::: Davenport, T. A. (2018). Nineteen and Up study (19Up): understanding pathways to mental health disorders in young Australian twins. BMJ Open, 8, e018959. Crowe, R. R., Noyes, R., Pauls, D. L., & Slymen, D. (1983). A family study of panic disorder. Archives of General Psychiatry , 40, 1065–1069. Department of Health. (2009). Prevalence of mental disorders in the Australian population. Retrieved from http://www.health.gov.au/internet/publications/ publishing.nsf/Content/mental-pubs-m-mhaust2-toc~mental-pubs-m- mhaust2-hig~mental-pubs-m-mhaust2-hig-pre Galea, S., Nandi, A., & Vlahov, D. (2005). The epidemiology of post-traumatic stress disorder after disasters. Epidemiologic Reviews, 27,7 8–91. Gavranidou, M., & Rosner, R. (2003). The weaker sex? Gender and post-trau- matic stress disorder. Depression and Anxiety , 17, 130 –139. Gillespie, N. A., Henders, A. K., Davenport, T. A., Hermens, D. F., Wright, M. J., Martin, N. G., & Hickie, I. B.(2013). The Brisbane Longitudinal Twin Study: Pathways to Cannabis Use, Abuse, and Dependence project — Current status, preliminary results, and future directions. Twin Research and Human Genetics , 16,2 1–33. Gillespie, N. A., Neale, M. C., Bates, T. C., Eyler, L. T., Fennema-Notestine, C., Vassileva, J., ::: Thompson, P. M. (2018). Testing associations between cannabis use and subcortical volumes in two large population-based samples. Addiction. Advance online publication. Gillespie, N. A., Neale, M. C., & Kendler, K. S. (2009). Pathways to cannabis abuse: A multi-stage model from cannabis availability, cannabis initiation and progression to abuse. Addiction, 104, 430–438. Gorman, J. M. (1996). Comorbid depression and anxiety spectrum disorders. Depression and Anxiety , 4, 160–168. Grant, J. E. (2014). Obsessive –compulsive disorder. New England Journal of Medicine, 371, 646–653. Hickie, I. B., Davenport, T. A., Hadzi-Paviovic, D., Koschera, A., Naismith, S. L., Scott, E. M., & Wilhelm, K. A. (2001). Development of a simple screening tool for common mental disorders in general practice. Medical Journal of Australia , 175, S10–S17. Hudson, J. I., Hiripi, E., Pope Jr, H. G., & Kessler, R. C. (2007). The preva- lence and correlates of eating disorders in the National Comorbidity Survey Replication. Biological Psychiatry, 61, 348 –358. Joel, D., Tarrasch, R., Berman, Z., Mukamel, M., & Ziv, E. (2014). Queering gender: Studying gender identity in ‘normative’ individuals. Psychology & Sexuality, 5, 291 –321. Johnson, S. L. (2004). Defining bipolar disorder. In S. L. Johnson & R. L. Leahy (Eds.), Psychological treatment of bipolar disorder (pp. 3–16). New York, NY: The Guilford Press. Karno, M., Golding, J. M., Sorenson, S. B., & Burnam, M. A. (1988). The epi- demiology of obsessive-compulsive disorder in five US communities. Archives of General Psychiatry , 45, 1094 –1099. Kessler, R. C., Nelson, C. B., McGonagle, K. A., Edlund, M. J., Frank, R. G., & Leaf, P. J. (1996). The epidemiology of co-occurring addictive and mental disorders: implications for prevention and service utilization. American Journal of Orthopsychiatry , 66,1 7–31. Krueger, R. F., Caspi, A., Moffitt, T. E., & Silva, P. A. (1998). The structure and stability of common mental disorders (DSM-III-R): A longitudinal-epi- demiological study. Journal of Abnormal Psychology , 107, 216–227. Laisk, T., Kukushkina, V., Palmer, D., Laber, S., Chen, C.-Y., Ferreira, T., ::: Smoller, J. W. (2018). GWAS meta-analysis highlights the hypotha- lamic-pituitary-gonadal axis (HPG axis) in the genetic regulation of men- strual cycle length. bioRxiv, 333708. Lawrence, D., Hafekost, J., Johnson, S. E., Saw, S., Buckingham, W. J., Sawyer, M. G., ::: Zubrick, S. R. (2016). Key findings from the second Australian Child and Adolescent Survey of Mental Health and Wellbeing. Australian & New Zealand Journal of Psychiatry , 50, 876–886. 162 Brittany L. Mitchell et al. https://doi.org/10.1017/thg.2019.27 Published online by Cambridge University Press Liddell, B. J., Nickerson, A., Sartor, L., Ivancic, L., & Bryant, R. A.(2016). The generational gap: mental disorder prevalence and disability amongst first and second generation immigrants in Australia. Journal of Psychiatric Research , 83, 103 –111. L´opez-Solà, C., Fontenelle, L. F., Alonso, P., Cuadras, D., Foley, D. L., Pantelis, C., ::: Soriano-Mas, C. (2014). Prevalence and heritability of obsessive-compulsive spectrum and anxiety disorder symptoms: A survey of the Australian Twin Registry. American Journal of Medical Genetics Part B: Neuropsychiatric Genetics , 165, 314–325. Maislin, G., Pack, A. I., Kribbs, N. B., Smith, P. L., Schwartz, A. R., Kline, L. R., ::: Dinges, D. F. (1995). A survey screen for prediction of apnea. Sleep, 18, 158 –166. Marshall, J. R. (1994). The diagnosis and treatment of social phobia and alcohol abuse. Bulletin of the Menninger Clinic , 58, A58–A66. Martin, N. G., Eaves, L. J., Kearsey, M. J., & Davies, P. (1978). The power of the classical twin study. Heredity, 40,9 7–116. Mbarek, H., Steinberg, S., Nyholt, D. R., Gordon, S. D., Miller, M. B., McRae, A. F., ::: De Geus, E. J. (2016). Identification of common genetic variants influencing spontaneous dizygotic twinning and female fertility. The American Journal of Human Genetics , 98, 898–908. Merikangas, K., Angst, J., Eaton, W., Canino, G., Rubio-Stipec, M., Wacker, H., ::: Whitaker, A. (1996). Comorbidity and boundaries of affective dis- orders with anxiety disorders and substance misuse: results of an international task force. The British Journal of Psychiatry , 168,5 8–67. Mitchison, D., & Hay, P. J. (2014). The epidemiology of eating disorders: Genetic, environmental, and societal factors. Clinical Epidemiology, 6,8 9–97. Myrick, H., & Brady, K. (2003). Current review of the comorbidity of affective, anxiety, and substance use disorders.Current Opinion in Psychiatry, 16,2 6 1–270. National Collaborating Centre for Mental Health . (2014). Psychosis and schizophrenia in adults: Treatment and management. NICE Clinical Guideline , 178,1 –59. Painter, J. N., Willemsen, G., Nyholt, D., Hoekstra, C., Duffy, D. L., Henders, A. K., ::: Skolnick, M. (2010). A genome wide linkage scan for dizygotic twinning in 525 families of mothers of dizygotic twins.Human Reproduction, 25, 1569–1580. Regier, D. A., Farmer, M. E., Rae, D. S., Locke, B. Z., Keith, S. J., Judd, L. L., & Goodwin, F. K. (1990). Comorbidity of mental disorders with alcohol and other drug abuse: Results from the Epidemiologic Catchment Area (ECA) study. JAMA, 264, 2511–2518. Schmaal, L., Veltman, D. J., van Erp, T. G., Sämann, P., Frodl, T., Jahanshad, N., ::: Niessen, W. (2016). Subcortical brain alterations in major depressive disorder: Findings from the ENIGMA Major Depressive Disorder Working Group. Molecular Psychiatry, 21, 806–812. Scott, J., Chant, D., Andrews, G., & McGrath, J. (2006). Psychotic-like expe- riences in the general community: The correlates of CIDI psychosis screen items in an Australian sample. Psychological Medicine, 36, 231–238. Thornton, L. M., Mazzeo, S. E., & Bulik, C. M. (2010). The heritability of eating disorders: Methods and current findings. Current Topics in Behavioral Neurosciences, 6, 141 –156. V i s s c h e r ,P .M . ,A n d r e w ,T . ,&N y h o l t ,D .R . (2008). Genome-wide association studies of quantitative traits with related individuals: Little (power) lost but much to be gained. European Journal of Human Genetics, 16, 387 –390. Volpe, U., Tortorella, A., Manchia, M., Monteleone, A. M., Albert, U., & Monteleone, P. (2016). Eating disorders: What age at onset? Psychiatry Research, 238, 225–227. Ward Jr., J. H. (1963). Hierarchical grouping to optimize an objective function. Journal of the American Statistical Association , 58, 236–244. Weissman, M. M., Bland, R. C., Canino, G. J., Faravelli, C., Greenwald, S., Hwu, H.-G., ::: Lellouch, J. (1997). The cross-national epidemiology of panic disorder. Archives of General Psychiatry , 54, 305 –309. Weissman, M. M., & Klerman, G. L. (1977). Sex differences and the epidemi- ology of depression. Archives of General Psychiatry , 34,9 8–111. Wright, M. J., & Martin, N. G. (2004). Brisbane Adolescent Twin Study: Outline of study methods and research projects. Australian Journal of Psychology, 56, 65–78. Twin Research and Human Genetics 163 https://doi.org/10.1017/thg.2019.27 Published online by Cambridge University Press