Introduction
Three quarters of major mental disorders emerge between childhood and young adulthood, with a peak onset during adolescence (Kessler et al., Reference Kessler, Berglund, Demler, Jin, Merikangas and Walters2005; Patel et al., Reference Patel, Flisher, Hetrick and McGorry2007). Thus, most mental illnesses first appear within a highly critical period of (neuro-) development (Arnett, Reference Arnett2004; Paus et al., Reference Paus, Keshavan and Giedd2008), before the most productive period of adult life. This can have a significant impact on the psychosocial trajectories of affected individuals and contributes considerably to the global economic burden of non-communicable disease (Patel et al., Reference Patel, Flisher, Hetrick and McGorry2007; Bloom et al., Reference Bloom, Cafiero, Jané-Llopis, Abrahams-Gessel, Bloom, Fathima, Feigl, Gaziano, Mowafi, Pandya, Prettner, Rosenberg, Seligman, Stein and Weinstein2011).
Identifying young people at risk of developing mental illness and taking action before the illness is well-established and potentially chronic is therefore an important goal. While this idea was proposed a long time ago (e.g. Meares, Reference Meares1959), the last two decades have witnessed a proliferation in psychiatric prediction research, predominantly in the area of psychosis. This was initiated by the formulation of a set of clinical criteria (termed ‘ultra-high’ or ‘clinical high’ risk criteria, UHR/CHR) which aim to identify help-seeking young people in the prodromal phase of a first psychotic episode. While the UHR approach has not been without its critics, it has led to a shift towards pre-emptive intervention for psychosis and more recently has facilitated the adoption of the clinical staging model in psychiatry.
In clinical staging models, disorder states are defined according to stages, ranging from a pre-symptomatic at-risk stage (stage 0, usually defined through genetic vulnerability), to undifferentiated general symptoms (stage 1a; e.g. anxiety, depression), to moderate and more specific symptoms alongside functional decline (stage 1b, ‘ultra-high risk’). Further stages include full-threshold (stage 2), more persistent (stage 3), and severe, unremitted illness (stage 4) (McGorry, Reference McGorry2007, Reference McGorry2013; McGorry et al., Reference McGorry, Nelson, Goldstone and Yung2010; McGorry et al., Reference McGorry, Keshavan, Goldstone, Amminger, Allott, Berk, Lavoie, Pantelis, Yung, Wood and Hickie2014a). Between stages 1 (subthreshold syndrome) and stages 2–4 (persistent full-threshold disorder), a step-like function (‘transition’) is assumed (McGorry et al., Reference McGorry, Keshavan, Goldstone, Amminger, Allott, Berk, Lavoie, Pantelis, Yung, Wood and Hickie2014a). Progression through stages (transition) is not assumed to be inevitable, but amelioration and remission are possible. Hence, the staging model takes a preventative approach, targeted at preventing disorder onset and/or progression, with treatment regimen selected according to stage (McGorry, Reference McGorry2007; McGorry et al., Reference McGorry, Nelson, Goldstone and Yung2010, Reference McGorry, Keshavan, Goldstone, Amminger, Allott, Berk, Lavoie, Pantelis, Yung, Wood and Hickie2014a). This approach is consistent with the current trend in psychiatry to think beyond traditional diagnostic categories, reflected in the National Institute of Mental Health (NIMH) Research Domain Criteria (RDoC) (Cuthbert and Insel, Reference Cuthbert and Insel2010, Reference Cuthbert and Insel2013; Insel et al., Reference Insel, Cuthbert, Garvey, Heinssen, Pine, Quinn, Sanislow and Wang2010), and more personalised medicine in psychiatry, with treatment tailored to stage of disorder and an individual profile of risk factors (Insel, Reference Insel2014).
An ongoing debate concerns the ‘splitting v. lumping’ issue in clinical staging models, i.e. whether to favour a general, transdiagnostic staging approach (‘lumping’) or a staging model per disorder (‘splitting’) (Duffy and Malhi, Reference Duffy and Malhi2017; Scott and Henry, Reference Scott and Henry2017). While the original model proposed a general, transdiagnostic staging approach, taking into account overlapping early clinical phenotypes (McGorry et al., Reference McGorry, Hickie, Yung, Pantelis and Jackson2006; McGorry, Reference McGorry2007; McGorry and Nelson, Reference McGorry and Nelson2016), separate staging models for each disorder have been suggested (Berk et al., Reference Berk, Conus, Lucas, Hallam, Malhi, Dodd, Yatham, Yung and McGorry2007a; Treasure et al., Reference Treasure, Stein and Maguire2015; Verduijn et al., Reference Verduijn, Milaneschi, van Hemert, Schoevers, Hickie, Penninx and Beekman2015). It is our view that the degree of theoretical discussion and empirical research between diagnostic silos (i.e. cross-diagnostic discourse) has been limited and is reflected, for example, in the disorder-specific focus of conferences and journals. Fostering such cross-diagnostic discourse is necessary to increase the knowledge base and assess the merits of a general, transdiagnostic staging model compared with disorder-specific staging models.
Therefore, the present scoping review (Arksey and O'Malley, Reference Arksey and O'Malley2005) aims to facilitate such cross-diagnostic discourse by (1) mapping the broad and diverse body of literature regarding at-risk approaches and identified antecedents for the onset of three major mental disorders [psychosis, bipolar disorder (BD) and depression], and (2) placing these in the context of clinical staging. Given the extensive scope of the research, we focus on prospective longitudinal studies of young at-risk populations (stages 0–1b) and identified clinical antecedents/predictors of homotypic disease progression (stage 2 exit syndromes). We focus on clinical (rather than cognitive, genetic or neurobiological) antecedents because of the scope of the literature as well as the emphasis to date on the definition of stages according to clinical variables.
Methods
As this review focuses on samples in which prevention is considered feasible, we included key longitudinal original research studies involving young people in enriched samples, i.e. at stages 0–1b at baseline, and identified clinical antecedents/predictors of homotypic illness progression. We excluded registry based studies, specific sub-populations, cross-sectional and retrospective studies; studies focusing on non-clinical predictors; studies that only assessed the course of the condition/rates of onset (i.e. no predictors identified); and studies investigating variables affecting symptom severity (rather than incidence of disorder). Incidence of disorder is operationalised according to DSM IV/5 or ICD-10 diagnoses as most studies to date have used this approach to determine outcome and stage transition, although we recognise that future studies of risk prediction may be less rooted in the DSM/ICD system (Hyman, Reference Hyman2010, Reference Hyman2011).
Pubmed and Google Scholar databases were screened for relevant literature using the following search terms: at-risk, high-risk, prospective, longitudinal, young people, adolescents, prediction, risk factors, antecedents, precursors, depression, psychosis, schizophrenia, mania, bipolar. Reviews and reference lists of identified articles were examined as well as articles citing the included studies using the Google scholar functionality ‘cited by’. The literature search was further enriched by the authors’ expert knowledge of the existing literature.
Results
The identified at-risk approaches will be described for each disorder separately (psychosis, BD, depression) in line with the predominantly disorder-specific prediction research published to date.
Psychosis
Familial high-risk (FHR) studies: stage 0
Traditionally, at-risk research in psychosis focused on FHR studies, i.e. the longitudinal study of first-degree relatives of affected patients, most typically the (not yet affected) offspring of one or two parents diagnosed with schizophrenia. In terms of clinical staging, this corresponds to stage 0: a pre-symptomatic vulnerability state. Genetic factors have been shown to be the strongest individual-level predictor of schizophrenia (Carter et al., Reference Carter, Schulsinger, Parnas, Cannon and Mednick2002; Sullivan, Reference Sullivan2005) and evidence from family studies points to a lifetime risk of psychosis of around 13% if one parent is affected and 46% if both parents are affected (Faraone et al., Reference Faraone, Taylor and Tsuang2002). The main aims of FHR studies are to identify rates of disorder onset (i.e. progression to stage 2) and premorbid predictors of adult illness (Cornblatt, Reference Cornblatt2002; Goldstein et al., Reference Goldstein, Buka, Seidman and Tsuang2010). The major first-generation FHR studies that followed offspring into adulthood involve the New York Infant HR study (Fish et al., Reference Fish, Marcus, Hans, Auerbach and Perdue1992), the Copenhagen High Risk study (Mednick and McNeil, Reference Mednick and McNeil1968), the Israeli HR study (Marcus et al., Reference Marcus, Hans, Nagler, Auerbach, Mirsky and Aubrey1987; Ingraham et al., Reference Ingraham, Kugelmass, Frenkel, Nathan and Mirsky1995), the New York HR study (Erlenmeyer-Kimling et al., Reference Erlenmeyer-Kimling, Squires-Wheeler, Adamo, Bassett, Cornblatt, Kestenbaum, Rock, Roberts and Gottesman1995), the Finnish Adoptive Family study (Tienari et al., Reference Tienari, Sorri, Lahti, Naarala, Wahlberg, Moring, Pohjola and Wynne1987) and the Helsinki HR study (Wrede et al., Reference Wrede, Mednick, Huttunen and Nilsson1980). The more recent, second-generation FHR studies include the Edinburgh HR study (Johnstone et al., Reference Johnstone, Abukmeil, Byrne, Clafferty, Grant, Hodges, Lawrie and Owens2000) and the Pittsburgh Risk Evaluation Program (Keshavan et al., Reference Keshavan, Diwadkar, Montrose, Stanley and Pettegrew2004). These studies aim to address weaknesses of the earlier cohorts, such as small sample sizes, and employ more rigorous methodologies. The Edinburgh HR study, started in 1994, involves offspring of at least two first- or second-degree affected relatives (Johnstone et al., Reference Johnstone, Abukmeil, Byrne, Clafferty, Grant, Hodges, Lawrie and Owens2000). Despite varying widely in methodology, these FHR studies have identified a range of clinical and psychosocial risk factors in children and young adults for progression to stage 2 psychosis.
Identified antecedents/predictors of progression to stage 2 psychosis
Findings from the Copenhagen HR study, the first study with a substantial sample size (N = 311), suggest that (i) high ratings on formal thought disorder and defective emotional contact (Parnas et al., Reference Parnas, Schulsinger, Schulsinger, Mednick and Teasdale1982) and (ii) high scores on unusual thoughts/experiences and psychoticism predicted onset of full-threshold, i.e. stage 2, schizophrenia 10 and 25 years later (Carter et al., Reference Carter, Parnas, Cannon, Schulsinger and Mednick1999). The Edinburgh HR study identified that the total score on schizotypy discriminated between those who did and did not transition to psychosis, with social anxiety and withdrawal identified as the strongest single predictors of transition to psychosis (Miller et al., Reference Miller, Byrne, Hodges, Lawrie, Owens and Johnstone2002a; Johnstone et al., Reference Johnstone, Ebmeier, Miller, Owens and Lawrie2005). Higher anxiety ratings at age 16 discriminated between those who later received a schizophrenia spectrum diagnosis (SSD) by age 26 and those who did not in the Israeli HR study (Kugelmass et al., Reference Kugelmass, Faber, Ingraham, Frenkel, Nathan, Mirsky and Ben Shakhar1995). Findings from the New York HR study indicated that anhedonia assessed at age 20 predicted psychosis in females (Erlenmeyer-Kimling et al., Reference Erlenmeyer-Kimling, Cornblatt, Rock, Roberts, Bell and West1993). The more recent Pittsburgh FHR study tested a new set of clinical criteria: higher scores on (i) positive symptoms and disorganisation and (ii) perceptual aberration/magical ideation were most predictive of transition (Tandon et al., Reference Tandon, Montrose, Shah, Rajarethinam, Diwadkar and Keshavan2012). Magical ideation, social anhedonia and perceptual aberration were directly related to later transition to psychosis (Shah et al., Reference Shah, Eack, Montrose, Tandon, Miewald, Prasad and Keshavan2012).
Several FHR studies identified that social adjustment difficulties during childhood and adolescence preceded the onset of psychosis. HR offspring who later developed schizophrenia had more disruptive school behaviour in the Copenhagen HR study (Olin et al., Reference Olin, John and Mednick1995; Carter et al., Reference Carter, Schulsinger, Parnas, Cannon and Mednick2002), a predictor previously identified in the Israeli HR studies for SSD (Marcus et al., Reference Marcus, Hans, Nagler, Auerbach, Mirsky and Aubrey1987), as well as greater pre-school social adjustment problems in the Helsinki HR study (Niemi et al., Reference Niemi, Suvisaari, Haukka and Lonnqvist2005). Childhood behavioural problems in those without substance abuse (Amminger et al., Reference Amminger, Pape, Rock, Roberts, Ott, Squires-Wheeler, Kestenbaum and Erlenmeyer-Kimling1999) as well as withdrawn and aggressive behaviours at ages 13–16 (Miller et al., Reference Miller, Byrne, Hodges, Lawrie and Johnstone2002b), predicted future psychosis in the New York HR and Edinburgh HR studies, respectively.
A number of FHR studies indicated a stressful rearing environment, such as institutionalisation, family instability, or poor relationships with parents, is a risk marker for development of psychosis in HR offspring (Tienari et al., Reference Tienari, Sorri, Lahti, Naarala, Wahlberg, Moring, Pohjola and Wynne1987; Cannon et al., Reference Cannon, Mednick and Parnas1990; Carter et al., Reference Carter, Schulsinger, Parnas, Cannon and Mednick2002; Schiffman et al., Reference Schiffman, LaBrie, Carter, Cannon, Schulsinger, Parnas and Mednick2002; Tienari et al., Reference Tienari, Wynne, Sorri, Lahti, Laksy, Moring, Naarala, Nieminen and Wahlberg2004). Recently, it has been shown that a polyenviomic risk score, comprising environmental measures such as urbanicity, various types of childhood adversity, cannabis use, etc. predicted progression to psychosis in young FHR individuals (Padmanabhan et al., Reference Padmanabhan, Shah, Tandon and Keshavan2017).
UHR and basic symptoms studies: stage 1 a/b
In the late 1990s, growing evidence demonstrated that the sooner the intervention for schizophrenia was introduced, the more favourable the outcome; similarly, the longer the duration of untreated psychosis, the poorer the outcome (Marshall et al., Reference Marshall, Lewis, Lockwood, Drake, Jones and Croudace2005). These findings suggested that intervening even earlier, before the development of a first episode of psychosis, would further enhance outcomes. Together with an increased awareness of the psychosis prodrome, the development of the UHR criteria introduced a different at-risk approach to the traditional FHR studies, focusing on early detection and intervention in help-seeking individuals (Yung et al., Reference Yung, McGorry, McFarlane, Jackson, Patton and Rakkar1996; Cornblatt, Reference Cornblatt2002). In contrast to the FHR studies, primarily interested in the identification of premorbid predictors of schizophrenia, UHR research focuses on the identification of young people in the putatively prodromal period of the disorder (hence the term ‘ultra’ HR), with the long-term goal of ameliorating, delaying or even preventing the onset of psychosis (Yung et al., Reference Yung, McGorry, McFarlane, Jackson, Patton and Rakkar1996; Yung et al., Reference Yung, Yuen, McGorry, Phillips, Kelly, Dell'Olio, Francey, Cosgrave, Killackey, Stanford, Godfrey and Buckby2005; Fusar-Poli et al., Reference Fusar-Poli, Borgwardt, Bechdolf, Addington, Riecher-Rossler, Schultze-Lutter, Keshavan, Wood, Ruhrmann, Seidman, Valmaggia, Cannon, Velthorst, De Haan, Cornblatt, Bonoldi, Birchwood, McGlashan, Carpenter, McGorry, Klosterkotter, McGuire and Yung2013). In terms of clinical staging, this at-risk approach would correspond to a stage 1b: help seeking young people with attenuated symptoms. As a complementary approach to the symptom-oriented UHR approach, the ‘basic symptom’ (BS) approach has focused on subtle, self-experienced disturbances in affect, cognition, speech, perception and motor action (Schultze-Lutter et al., Reference Schultze-Lutter, Ruhrmann, Klosterkotter, Johanessen, Martindale and Culberg2006; Schultze-Lutter et al., Reference Schultze-Lutter, Klosterkotter, Picker, Steinmeyer and Ruhrmann2007). The BS approach putatively identifies young people at an earlier stage than the UHR approach, and may therefore be conceptualised in terms of clinical staging as 1a (Klosterkotter, Reference Klosterkotter2008).
UHR and BS studies generally take the form of a baseline assessment and a follow-up period of 6–24 months, with several studies extending this period to the medium (2–3 years) and longer term (up to 15 years) (Nelson et al., Reference Nelson, Yuen, Wood, Lin, Spiliotacopoulos, Bruxner, Broussard, Simmons, Foley, Brewer, Francey, Amminger, Thompson, McGorry and Yung2013). While these UHR/BS approaches come from an early intervention perspective, a related goal has been to identify risk markers for the illness (Yung et al., Reference Yung, McGorry, McFarlane, Jackson, Patton and Rakkar1996). Although the UHR studies vary in methodology and the domains of variables investigated, several clinical and psychosocial risk markers have been identified. Main cohorts include the North American Prodrome Longitudinal study (NAPLS and NAPLS 2) (Addington et al., Reference Addington, Cadenhead, Cannon, Cornblatt, McGlashan, Perkins, Seidman, Tsuang, Walker, Woods and Heinssen2007; Addington et al., Reference Addington, Cadenhead, Cornblatt, Mathalon, McGlashan, Perkins, Seidman, Tsuang, Walker, Woods, Addington and Cannon2012), the PACE 400 cohort (Nelson et al., Reference Nelson, Yuen, Wood, Lin, Spiliotacopoulos, Bruxner, Broussard, Simmons, Foley, Brewer, Francey, Amminger, Thompson, McGorry and Yung2013), the Dutch Prediction of Psychosis study (DUPS) (Velthorst et al., Reference Velthorst, Nieman, Becker, van de Fliert, Dingemans, Klaassen, de Haan, van Amelsvoort and Linszen2009), the European Prediction of Psychosis study (EPOS) (Ruhrmann et al., Reference Ruhrmann, Schultze-Lutter, Salokangas, Heinimaa, Linszen, Dingemans, Birchwood, Patterson, Juckel, Heinz, Morrison, Lewis, von Reventlow and Klosterkotter2010) and the Outreach and Support in South London (OASIS) cohort (Fusar-Poli et al., Reference Fusar-Poli, Byrne, Valmaggia, Day, Tabraham, Johns, McGuire and Team2010).
Identified antecedents/predictors of progression to stage 2 psychosis
A large number of UHR studies identified the severity of (subthreshold) positive psychotic symptoms at baseline as a predictor of transition to stage 2 (i.e. full psychosis) (Lencz et al., Reference Lencz, Smith, Auther, Correll and Cornblatt2003; Yung et al., Reference Yung, Phillips, Yuen, Francey, McFarlane, Hallgren and McGorry2003; Haroun et al., Reference Haroun, Dunn, Haroun and Cadenhead2006; Cannon et al., Reference Cannon, Cadenhead, Cornblatt, Woods, Addington, Walker, Seidman, Perkins, Tsuang, McGlashan and Heinssen2008; Lemos-Giraldez et al., Reference Lemos-Giraldez, Vallina-Fernandez, Fernandez-Iglesias, Vallejo-Seco, Fonseca-Pedrero, Paino-Pineiro, Sierra-Baigrie, Garcia-Pelayo, Pedrejon-Molino, Alonso-Bada, Gutierrez-Perez and Ortega-Ferrandez2009; Ruhrmann et al., Reference Ruhrmann, Schultze-Lutter, Salokangas, Heinimaa, Linszen, Dingemans, Birchwood, Patterson, Juckel, Heinz, Morrison, Lewis, von Reventlow and Klosterkotter2010; Ziermans et al., Reference Ziermans, de Wit, Schothorst, Sprong, van Engeland, Kahn and Durston2014; Hengartner et al., Reference Hengartner, Heekeren, Dvorsky, Walitza, Rossler and Theodoridou2017). Constructs related to thought disorder/disorganisation (Klosterkotter et al., Reference Klosterkotter, Hellmich, Steinmeyer and Schultze-Lutter2001; Haroun et al., Reference Haroun, Dunn, Haroun and Cadenhead2006; Demjaha et al., Reference Demjaha, Valmaggia, Stahl, Byrne and McGuire2012; Nelson et al., Reference Nelson, Yuen, Wood, Lin, Spiliotacopoulos, Bruxner, Broussard, Simmons, Foley, Brewer, Francey, Amminger, Thompson, McGorry and Yung2013; DeVylder et al., Reference DeVylder, Muchomba, Gill, Ben-David, Walder, Malaspina and Corcoran2014; Addington et al., Reference Addington, Liu, Buchy, Cadenhead, Cannon, Cornblatt, Perkins, Seidman, Tsuang, Walker, Woods, Bearden, Mathalon and McGlashan2015; Cornblatt et al., Reference Cornblatt, Carrion, Auther, McLaughlin, Olsen, John and Correll2015; Brucato et al., Reference Brucato, Masucci, Arndt, Ben-David, Colibazzi, Corcoran, Crumbley, Crump, Gill, Kimhy, Lister, Schobel, Yang, Lieberman and Girgis2017) and unusual thought content, such as paranoid thoughts (Lencz et al., Reference Lencz, Smith, Auther, Correll and Cornblatt2003; Haroun et al., Reference Haroun, Dunn, Haroun and Cadenhead2006; Cannon et al., Reference Cannon, Cadenhead, Cornblatt, Woods, Addington, Walker, Seidman, Perkins, Tsuang, McGlashan and Heinssen2008; Thompson et al., Reference Thompson, Nelson and Yung2011; Addington et al., Reference Addington, Liu, Buchy, Cadenhead, Cannon, Cornblatt, Perkins, Seidman, Tsuang, Walker, Woods, Bearden, Mathalon and McGlashan2015; Brucato et al., Reference Brucato, Masucci, Arndt, Ben-David, Colibazzi, Corcoran, Crumbley, Crump, Gill, Kimhy, Lister, Schobel, Yang, Lieberman and Girgis2017), appear to be particularly associated with psychosis onset. Bizarre thinking (Velthorst et al., Reference Velthorst, Nieman, Becker, van de Fliert, Dingemans, Klaassen, de Haan, van Amelsvoort and Linszen2009; Ruhrmann et al., Reference Ruhrmann, Schultze-Lutter, Salokangas, Heinimaa, Linszen, Dingemans, Birchwood, Patterson, Juckel, Heinz, Morrison, Lewis, von Reventlow and Klosterkotter2010) and schizotypal personality disorder (Mason et al., Reference Mason, Startup, Halpin, Schall, Conrad and Carr2004; Ruhrmann et al., Reference Ruhrmann, Schultze-Lutter, Salokangas, Heinimaa, Linszen, Dingemans, Birchwood, Patterson, Juckel, Heinz, Morrison, Lewis, von Reventlow and Klosterkotter2010) have also been identified as predicting transition. Evidence for perceptual abnormalities as a predictor of transition is scarce (Mason et al., Reference Mason, Startup, Halpin, Schall, Conrad and Carr2004).
Predictors in the domain of negative symptoms have also been investigated. UHR patients high in negative symptoms (Amminger et al., Reference Amminger, Leicester, Yung, Phillips, Berger, Francey, Yuen and McGorry2006; Lemos-Giraldez et al., Reference Lemos-Giraldez, Vallina-Fernandez, Fernandez-Iglesias, Vallejo-Seco, Fonseca-Pedrero, Paino-Pineiro, Sierra-Baigrie, Garcia-Pelayo, Pedrejon-Molino, Alonso-Bada, Gutierrez-Perez and Ortega-Ferrandez2009; Demjaha et al., Reference Demjaha, Valmaggia, Stahl, Byrne and McGuire2012; Piskulic et al., Reference Piskulic, Addington, Cadenhead, Cannon, Cornblatt, Heinssen, Perkins, Seidman, Tsuang, Walker, Woods and McGlashan2012; Schlosser et al., Reference Schlosser, Jacobson, Chen, Sugar, Niendam, Li, Bearden and Cannon2012; Nelson et al., Reference Nelson, Yuen, Wood, Lin, Spiliotacopoulos, Bruxner, Broussard, Simmons, Foley, Brewer, Francey, Amminger, Thompson, McGorry and Yung2013; Valmaggia et al., Reference Valmaggia, Stahl, Yung, Nelson, Fusar-Poli, McGorry and McGuire2013), particularly in the domains of amotivation as defined by avolition/apathy or anhedonia (Yung et al., Reference Yung, Phillips, Yuen, Francey, McFarlane, Hallgren and McGorry2003; Mason et al., Reference Mason, Startup, Halpin, Schall, Conrad and Carr2004; Valmaggia et al., Reference Valmaggia, Stahl, Yung, Nelson, Fusar-Poli, McGorry and McGuire2013), alogia (Valmaggia et al., Reference Valmaggia, Stahl, Yung, Nelson, Fusar-Poli, McGorry and McGuire2013) and social isolation/withdrawal (Mason et al., Reference Mason, Startup, Halpin, Schall, Conrad and Carr2004; Velthorst et al., Reference Velthorst, Nieman, Becker, van de Fliert, Dingemans, Klaassen, de Haan, van Amelsvoort and Linszen2009; Piskulic et al., Reference Piskulic, Addington, Cadenhead, Cannon, Cornblatt, Heinssen, Perkins, Seidman, Tsuang, Walker, Woods and McGlashan2012; Brucato et al., Reference Brucato, Masucci, Arndt, Ben-David, Colibazzi, Corcoran, Crumbley, Crump, Gill, Kimhy, Lister, Schobel, Yang, Lieberman and Girgis2017), predict onset of stage 2 psychosis.
Related to the latter, poor functioning is one of the most consistently identified predictors of transition to psychosis (Yung et al., Reference Yung, Phillips, Yuen and McGorry2004; Amminger et al., Reference Amminger, Leicester, Yung, Phillips, Berger, Francey, Yuen and McGorry2006; Lam et al., Reference Lam, Hung and Chen2006; Yung et al., Reference Yung, Stanford, Cosgrave, Killackey, Phillips, Nelson and McGorry2006; Lemos-Giraldez et al., Reference Lemos-Giraldez, Vallina-Fernandez, Fernandez-Iglesias, Vallejo-Seco, Fonseca-Pedrero, Paino-Pineiro, Sierra-Baigrie, Garcia-Pelayo, Pedrejon-Molino, Alonso-Bada, Gutierrez-Perez and Ortega-Ferrandez2009; Velthorst et al., Reference Velthorst, Nieman, Becker, van de Fliert, Dingemans, Klaassen, de Haan, van Amelsvoort and Linszen2009; Ruhrmann et al., Reference Ruhrmann, Schultze-Lutter, Salokangas, Heinimaa, Linszen, Dingemans, Birchwood, Patterson, Juckel, Heinz, Morrison, Lewis, von Reventlow and Klosterkotter2010; Velthorst et al., Reference Velthorst, Nieman, Linszen, Becker, de Haan, Dingemans, Birchwood, Patterson, Salokangas, Heinimaa, Heinz, Juckel, von Reventlow, French, Stevens, Schultze-Lutter, Klosterkotter and Ruhrmann2010; Dragt et al., Reference Dragt, Nieman, Veltman, Becker, van de Fliert, de Haan and Linszen2011; Thompson et al., Reference Thompson, Nelson and Yung2011; Nelson et al., Reference Nelson, Yuen, Wood, Lin, Spiliotacopoulos, Bruxner, Broussard, Simmons, Foley, Brewer, Francey, Amminger, Thompson, McGorry and Yung2013). In more recent studies, poor social functioning and social adjustment in particular have been identified as risk factors for transition (Cannon et al., Reference Cannon, Cadenhead, Cornblatt, Woods, Addington, Walker, Seidman, Perkins, Tsuang, McGlashan and Heinssen2008; Fusar-Poli et al., Reference Fusar-Poli, Byrne, Valmaggia, Day, Tabraham, Johns, McGuire and Team2010; Velthorst et al., Reference Velthorst, Nieman, Linszen, Becker, de Haan, Dingemans, Birchwood, Patterson, Salokangas, Heinimaa, Heinz, Juckel, von Reventlow, French, Stevens, Schultze-Lutter, Klosterkotter and Ruhrmann2010; Cornblatt et al., Reference Cornblatt, Carrion, Addington, Seidman, Walker, Cannon, Cadenhead, McGlashan, Perkins, Tsuang, Woods, Heinssen and Lencz2012; Schlosser et al., Reference Schlosser, Jacobson, Chen, Sugar, Niendam, Li, Bearden and Cannon2012; Tarbox et al., Reference Tarbox, Addington, Cadenhead, Cannon, Cornblatt, Perkins, Seidman, Tsuang, Walker, Heinssen, McGlashan and Woods2013; Valmaggia et al., Reference Valmaggia, Stahl, Yung, Nelson, Fusar-Poli, McGorry and McGuire2013; Nieman et al., Reference Nieman, Ruhrmann, Dragt, Soen, van Tricht, Koelman, Bour, Velthorst, Becker, Weiser, Linszen and de Haan2014; Cornblatt et al., Reference Cornblatt, Carrion, Auther, McLaughlin, Olsen, John and Correll2015).
Other identified clinical risk factors consist of depression and/or anxiety (Yung et al., Reference Yung, Phillips, Yuen, Francey, McFarlane, Hallgren and McGorry2003; Yung et al., Reference Yung, Phillips, Yuen and McGorry2004; Amminger et al., Reference Amminger, Leicester, Yung, Phillips, Berger, Francey, Yuen and McGorry2006), self-disturbance (Nelson et al., Reference Nelson, Thompson and Yung2012) and first rank symptoms (Nelson et al., Reference Nelson, Thompson and Yung2012; Morcillo et al., Reference Morcillo, Stochl, Russo, Zambrana, Ratnayake, Jones and Perez2015), sleep disturbance (Ruhrmann et al., Reference Ruhrmann, Schultze-Lutter, Salokangas, Heinimaa, Linszen, Dingemans, Birchwood, Patterson, Juckel, Heinz, Morrison, Lewis, von Reventlow and Klosterkotter2010), higher genetic loading due to family history (Cannon et al., Reference Cannon, Cadenhead, Cornblatt, Woods, Addington, Walker, Seidman, Perkins, Tsuang, McGlashan and Heinssen2008; Thompson et al., Reference Thompson, Nelson and Yung2011), early onset of psychiatric symptoms (Amminger et al., Reference Amminger, Leicester, Yung, Phillips, Berger, Francey, Yuen and McGorry2006) and substance abuse (Haroun et al., Reference Haroun, Dunn, Haroun and Cadenhead2006; Kristensen and Cadenhead, Reference Kristensen and Cadenhead2007; Cannon et al., Reference Cannon, Cadenhead, Cornblatt, Woods, Addington, Walker, Seidman, Perkins, Tsuang, McGlashan and Heinssen2008). Finally, a long duration of symptoms before first clinical contact has been identified as a predictor of transition in the PACE 400 group (Yung et al., Reference Yung, Phillips, Yuen and McGorry2004; Nelson et al., Reference Nelson, Yuen, Wood, Lin, Spiliotacopoulos, Bruxner, Broussard, Simmons, Foley, Brewer, Francey, Amminger, Thompson, McGorry and Yung2013; Nelson et al., Reference Nelson, Yuen, Lin, Wood, McGorry, Hartmann and Yung2016).
Less research has focused on psychosocial predictors. Among these are childhood trauma, particularly sexual trauma (Bechdolf et al., Reference Bechdolf, Thompson, Nelson, Cotton, Simmons, Amminger, Leicester, Francey, McNab, Krstev, Sidis, McGorry and Yung2010b; Thompson et al., Reference Thompson, Nelson, Yuen, Lin, Amminger, McGorry, Wood and Yung2014), lower education level (Ruhrmann et al., Reference Ruhrmann, Schultze-Lutter, Salokangas, Heinimaa, Linszen, Dingemans, Birchwood, Patterson, Juckel, Heinz, Morrison, Lewis, von Reventlow and Klosterkotter2010), receiving social security benefits (Dragt et al., Reference Dragt, Nieman, Veltman, Becker, van de Fliert, de Haan and Linszen2011) and perceived discrimination (Stowkowy et al., Reference Stowkowy, Liu, Cadenhead, Cannon, Cornblatt, McGlashan, Perkins, Seidman, Tsuang, Walker, Woods, Bearden, Mathalon and Addington2016).
Other enriched cohorts: stage 1
Several other at-risk approaches have also been developed, the most well-known being the psychometric psychosis-proneness approach of Chapman and Chapman, as adopted in the Wisconsin Psychosis Proneness Project in university students (Chapman and Chapman, Reference Chapman and Chapman1987; Chapman et al., Reference Chapman, Chapman, Kwapil, Eckblad and Zinser1994). Chapman's at-risk criteria consisted of individuals in the high-risk age group experiencing psychotic-like experiences, measured using a specifically developed interview (Chapman and Chapman, Reference Chapman and Chapman1980). Over a 10-year period, psychotic-like experiences, particularly perceptual aberrations, magical ideation and social anhedonia, but not physical anhedonia or impulsive non-conformity, predicted onset of psychotic disorder (Chapman et al., Reference Chapman, Chapman, Kwapil, Eckblad and Zinser1994; Kwapil, Reference Kwapil1998).
To summarise, prospective at-risk approaches in psychosis mainly comprise stage 0 (familial at risk) and UHR (stage 1b) conceptualisations and have identified a variety of clinical predictors for progression to stage 2 psychosis. Complementary to familial vulnerability and attenuated psychotic symptoms (inherent to at-risk designs), four domains of predictors for psychosis can be distilled: more severe attenuated psychotic symptoms (particularly in the domains of conceptual disorganisation and unusual thought content), poor functioning (particularly in the social domain and including behavioural problems), negative symptoms (especially anhedonia) and anxiety/depression.
Bipolar disorder
FHR studies: stage 0
Similar to psychosis research, traditional at-risk approaches in BD involved FHR approaches, as a positive family history constitutes the most robust risk factor predicting BD onset. Adolescent offspring from affected parents are at increased risk of developing BD (Tsuang and Faraone, Reference Tsuang and Faraone1990; Goes, Reference Goes2016) with lifetime estimates of BD I or II ranging between 6.9% (Egeland et al., Reference Egeland, Endicott, Hostetter, Allen, Pauls and Shaw2012) and 16% (Akiskal et al., Reference Akiskal, Downs, Jordan, Watson, Daugherty and Pruitt1985) in this group. The main aim of these studies was to identify risk factors and early indicators of emerging BD (Duffy et al., Reference Duffy, Alda, Crawford, Milin and Grof2007). Completed longitudinal FHR studies involving offspring of BD parents (or siblings of affected children) comprise the Memphis BD offspring study (Akiskal et al., Reference Akiskal, Downs, Jordan, Watson, Daugherty and Pruitt1985), the NIMH child rearing study (Radke-Yarrow, Reference Radke-Yarrow1998) and the Amish BD study (Egeland et al., Reference Egeland, Shaw, Endicott, Pauls, Allen, Hostetter and Sussex2003; Shaw et al., Reference Shaw, Egeland, Endicott, Allen and Hostetter2005). Major ongoing BD FHR studies include the Canadian BD risk cohort (Duffy et al., Reference Duffy, Alda, Crawford, Milin and Grof2007), the Dutch BD Offspring study (Hillegers et al., Reference Hillegers, Reichart, Wals, Verhulst, Ormel and Nolen2005) and the Pittsburgh Bipolar Offspring or BIOS study (Birmaher et al., Reference Birmaher, Axelson, Monk, Kalas, Goldstein, Hickey, Obreja, Ehmann, Iyengar, Shamseddeen, Kupfer and Brent2009). Another multi-site study involving sites in the USA and Australia is also currently underway (Nurnberger et al., Reference Nurnberger, McInnis, Reich, Kastelic, Wilcox, Glowinski, Mitchell, Fisher, Erpe, Gershon, Berrettini, Laite, Schweitzer, Rhoadarmer, Coleman, Cai, Azzouz, Liu, Kamali, Brucksch and Monahan2011).
Identified predictors of progression to stage 2 BD
For current purposes stage 2 BD is defined as onset of BD I or II, although we recognise that there is some ongoing debate in the field about exactly how stage 2 BD should be defined (Scott et al., Reference Scott, Leboyer, Hickie, Berk, Kapczinski, Frank, Kupfer and McGorry2013; Bechdolf et al., Reference Bechdolf, Ratheesh, Cotton, Nelson, Chanen, Betts, Bingmann, Yung, Berk and McGorry2014; Scott and Henry, Reference Scott and Henry2017). One of the identified clinical precursors of stage 2 BD (Hillegers et al., Reference Hillegers, Reichart, Wals, Verhulst, Ormel and Nolen2005; Reichart et al., Reference Reichart, van der Ende, Wals, Hillegers, Nolen, Ormel and Verhulst2005; Frankland et al., Reference Frankland, Roberts, Holmes-Preston, Perich, Levy, Lenroot, Hadzi-Pavlovic, Breakspear and Mitchell2017) and BD Spectrum (Hafeman et al., Reference Hafeman, Merranko, Axelson, Goldstein, Goldstein, Monk, Hickey, Sakolsky, Diler, Iyengar, Brent, Kupfer and Birmaher2016) comprise (sub)-threshold depression. The majority of FHR cohorts indicate that one or several depressive episodes may occur before the onset of a first (hypo) manic episode. Specifically, in a recent sample, psychomotor retardation and having had more than four depressive episodes predicted progression to stage 2 BD (Frankland et al., Reference Frankland, Roberts, Holmes-Preston, Perich, Levy, Lenroot, Hadzi-Pavlovic, Breakspear and Mitchell2017). Increasing evidence suggests that subthreshold (hypo-) manic symptoms predict progression to BD or BD spectrum (Axelson et al., Reference Axelson, Goldstein, Goldstein, Monk, Yu, Hickey, Sakolsky, Diler, Hafeman, Merranko, Iyengar, Brent, Kupfer and Birmaher2015; Hafeman et al., Reference Hafeman, Merranko, Axelson, Goldstein, Goldstein, Monk, Hickey, Sakolsky, Diler, Iyengar, Brent, Kupfer and Birmaher2016; Hafeman et al., Reference Hafeman, Merranko, Goldstein, Axelson, Goldstein, Monk, Hickey, Sakolsky, Diler, Iyengar, Brent, Kupfer, Kattan and Birmaher2017; Mesman et al., Reference Mesman, Nolen, Keijsers and Hillegers2017), generally in the period proximal to transition. More distal clinical antecedents of BD spectrum include anxiety and sleep disturbance (Duffy et al., Reference Duffy, Alda, Crawford, Milin and Grof2007; Egeland et al., Reference Egeland, Endicott, Hostetter, Allen, Pauls and Shaw2012; Hafeman et al., Reference Hafeman, Merranko, Axelson, Goldstein, Goldstein, Monk, Hickey, Sakolsky, Diler, Iyengar, Brent, Kupfer and Birmaher2016; Hafeman et al., Reference Hafeman, Merranko, Goldstein, Axelson, Goldstein, Monk, Hickey, Sakolsky, Diler, Iyengar, Brent, Kupfer, Kattan and Birmaher2017). Circadian/sleep disturbance, such as frequent awakenings during the night, has also been found to predict transition to BD (Levenson et al., Reference Levenson, Axelson, Merranko, Angulo, Goldstein, Mullin, Goldstein, Brent, Diler, Hickey, Monk, Sakolsky, Kupfer and Birmaher2015). Further predictors include greater mood lability/emotional disturbance (Akiskal et al., Reference Akiskal, Downs, Jordan, Watson, Daugherty and Pruitt1985; Meyer et al., Reference Meyer, Carlson, Youngstrom, Ronsaville, Martinez, Gold, Hakak and Radke-Yarrow2009; Egeland et al., Reference Egeland, Endicott, Hostetter, Allen, Pauls and Shaw2012; Hafeman et al., Reference Hafeman, Merranko, Axelson, Goldstein, Goldstein, Monk, Hickey, Sakolsky, Diler, Iyengar, Brent, Kupfer and Birmaher2016; Hafeman et al., Reference Hafeman, Merranko, Goldstein, Axelson, Goldstein, Monk, Hickey, Sakolsky, Diler, Iyengar, Brent, Kupfer, Kattan and Birmaher2017), poor psychosocial functioning and age of onset of mood disorder in the parent (Hafeman et al., Reference Hafeman, Merranko, Goldstein, Axelson, Goldstein, Monk, Hickey, Sakolsky, Diler, Iyengar, Brent, Kupfer, Kattan and Birmaher2017). Another study indicated that higher levels of exposure to maternal negativity (e.g. irritability, use of strong verbal control) predicts onset of BD in offspring of mothers with BD (Meyer et al., Reference Meyer, Carlson, Wiggs, Ronsaville, Martinez, Klimes-Dougan, Gold and Radke-Yarrow2006).
Based on available evidence, Duffy et al., proposed a staging model for BD, with non-specific symptoms (i.e. sleep/anxiety) appearing first, followed by depressive symptoms, finally leading to a first (hypo)-manic episode (Duffy et al., Reference Duffy, Alda, Hajek, Sherry and Grof2010). Recently, a risk calculator for BD spectrum was developed, comprising mood and anxiety, psychosocial functioning and age of onset of mood disorder in the parent as predictors, which discriminated (AUC 0.76) between those who developed BD spectrum disorder within a 5-year follow-up time window and those who did not (Hafeman et al., Reference Hafeman, Merranko, Goldstein, Axelson, Goldstein, Monk, Hickey, Sakolsky, Diler, Iyengar, Brent, Kupfer, Kattan and Birmaher2017). Although their definition of BP spectrum disorder including bipolar Not Otherwise Specified (NOS) cases, which would not meet criteria for stage 2 disorder, the findings held when these cases were removed and the model was applied only to BD I and II cases at follow-up point.
UHR studies: Stage 1b
The field of BD research has seen growing emphasis on early intervention and prevention, requiring the ability to identify young people in the prodrome of the disorder. Informed by the UHR concept, different sets of criteria/scales to identify young people at imminent risk of onset of BD have been proposed (Brietzke et al., Reference Brietzke, Mansur, Soczynska, Kapczinski, Bressan and McIntyre2012; Leopold et al., Reference Leopold, Ritter, Correll, Marx, Ozgurdal, Juckel, Bauer and Pfennig2012; Correll et al., Reference Correll, Olvet, Auther, Hauser, Kishimoto, Carrion, Snyder and Cornblatt2014). The Bipolar At Risk (BAR) criteria have been preliminarily tested; however, not yet robustly tested in sufficiently large prediction studies (Bechdolf et al., Reference Bechdolf, Nelson, Cotton, Chanen, Thompson, Kettle, Conus, Amminger, Yung, Berk and McGorry2010a; Scott et al., Reference Scott, Leboyer, Hickie, Berk, Kapczinski, Frank, Kupfer and McGorry2013; Bechdolf et al., Reference Bechdolf, Ratheesh, Cotton, Nelson, Chanen, Betts, Bingmann, Yung, Berk and McGorry2014; Scott et al., Reference Scott, Marwaha, Ratheesh, Macmillan, Yung, Morriss, Hickie and Bechdolf2017). Similar to the UHR criteria for psychosis, the BAR criteria consist of three at-risk groups: (1) sub-threshold mania, (2) depression and cyclothymic features, (3) depression and genetic risk of BD. Initial data suggest a transition rate to BD I or II of about 15–20% within 1 year in those who meet BAR criteria (Bechdolf et al., Reference Bechdolf, Nelson, Cotton, Chanen, Thompson, Kettle, Conus, Amminger, Yung, Berk and McGorry2010a; Bechdolf et al., Reference Bechdolf, Ratheesh, Cotton, Nelson, Chanen, Betts, Bingmann, Yung, Berk and McGorry2014).
Identified antecedents/predictors of progression to stage 2 BD
Identified clinical predictors include atypical depressive symptoms (anergia/hypersomnia) and antidepressant-emergent elation (Scott et al., Reference Scott, Marwaha, Ratheesh, Macmillan, Yung, Morriss, Hickie and Bechdolf2017). Other predictors using the BAR approach have not yet been investigated in longitudinal studies to date.
Other enriched cohorts: stage 1a/b
Another approach to predict BD onset has been the examination of cohorts with major depressive disorders at baseline. This approach may be of value given that depressive episodes are the most common polarity of onset among persons with BD (Etain et al., Reference Etain, Lajnef, Bellivier, Mathieu, Raust, Cochet, Gard, M'Bailara, Kahn, Elgrabli, Cohen, Jamain, Vieta, Leboyer and Henry2012), with an earlier age of onset than manic episodes (Berk et al., Reference Berk, Dodd, Callaly, Berk, Fitzgerald, de Castella, Filia, Filia, Tahtalian, Biffin, Kelin, Smith, Montgomery and Kulkarni2007b). Further approaches involve the selection of young people based on the presence of a BD NOS diagnosis (Axelson et al., Reference Axelson, Birmaher, Strober, Goldstein, Ha, Gill, Goldstein, Yen, Hower, Hunt, Liao, Iyengar, Dickstein, Kim, Ryan, Frankel and Keller2011; Alloy et al., Reference Alloy, Urosevic, Abramson, Jager-Hyman, Nusslock, Whitehouse and Hogan2012b), or symptom-enriched community samples (Angst et al., Reference Angst, Gamma and Endrass2003). Based on high sensitivity of the behavioural approach system in university students (Alloy et al., Reference Alloy, Bender, Whitehouse, Wagner, Liu, Grant, Jager-Hyman, Molz, Choi, Harmon-Jones and Abramson2012a), a study in a ‘behavioural high risk’ sample found that higher scores on the fun-seeking sub-scale of the Behavioural Activation Scale (Alloy et al., Reference Alloy, Urosevic, Abramson, Jager-Hyman, Nusslock, Whitehouse and Hogan2012b) and low social rhythm (i.e. patterns of daily regular structure of activities) as a risk factor for transition to BD, mainly BD II (Alloy et al., Reference Alloy, Boland, Ng, Whitehouse and Abramson2015). One other university student cohort has indicated the possible predictive value of the ‘Hypomanic Personality Scale’ (HPS) (Chapman et al., Reference Chapman, Chapman and Miller1982). Students identified to be at higher risk based on this instrument had a greater risk of developing manic or hypomanic episodes (Kwapil et al., Reference Kwapil, Miller, Zinser, Chapman, Chapman and Eckblad2000); however, the HPS did not identify the risk of incident DSM-IV BD in a later sample (Walsh et al., Reference Walsh, DeGeorge, Barrantes-Vidal and Kwapil2015).
While stage 1a and b approaches have been developed, the main at-risk approach to date has been the FHR (i.e. stage 0) approach. Most identified predictors emerged from the FHR studies and have included three main predictors (in addition to genetic predisposition): (sub) threshold depression, sleep disturbance and anxiety (more distal predictors), general mood lability and subthreshold hypomanic symptoms (presumably more prodromal).
Depression
FHR studies: stage 0
As with psychosis and mania, important prediction research in depression has been based on an FHR, i.e. stage 0 approach. Having a depressed parent constitutes one of the strongest risk factors for subsequent MDD. Offspring of depressed parents are at three- to fivefold increased risk of depression or other serious mental illness (Radke-Yarrow et al., Reference Radke-Yarrow, Nottelmann, Martinez, Fox and Belmont1992; Weissman et al., Reference Weissman, Warner, Wickramaratne, Moreau and Olfson1997; Lieb et al., Reference Lieb, Isensee, Hofler, Pfister and Wittchen2002; Weissman et al., Reference Weissman, Wickramaratne, Nomura, Warner, Pilowsky and Verdeli2006; Weissman et al., Reference Weissman, Wickramaratne, Gameroff, Warner, Pilowsky, Kohad, Verdeli, Skipper and Talati2016) with more than half being affected by depression by the age of 20 (Weissman et al., Reference Weissman, Fendrich, Warner and Wickramaratne1992). The major MDD FHR cohorts, which show some overlap with the FHR cohorts for BD, comprise the UCLA Family Stress Project (Hammen et al., Reference Hammen, Burge, Burney and Adrian1990), Yale Family Study of Major depression (Fendrich et al., Reference Fendrich, Warner and Weissman1990; Weissman et al., Reference Weissman, Wickramaratne, Gameroff, Warner, Pilowsky, Kohad, Verdeli, Skipper and Talati2016), the Early Prediction of Adolescent Depression study (Mars et al., Reference Mars, Collishaw, Smith, Thapar, Potter, Sellers, Harold, Craddock, Rice and Thapar2012) and the Binghamton HR study (Gibb et al., Reference Gibb, Grassia, Stone, Uhrlass and McGeary2012). Similar to psychosis risk, these longitudinal studies aim to identify risk factors and mechanisms for the onset of MDD in offspring, generally with the long-term goal of developing interventions to prevent the first episode of MDD (Gotlib et al., Reference Gotlib, Joormann and Foland-Ross2014).
Identified antecedents/predictors of progression to stage 2 depression
While the current review is based on the published studies to date, we note that most studies have taken DSM or ICD-defined MDD as the outcome of interest. However, clinical staging models propose that stage 2 depression would more accurately be represented by moderate to severe cases of MDD (Hickie et al., Reference Hickie, Scott, Hermens, Naismith, Guastella, Kaur, Sidis, Whitwell, Glozier, Davenport, Pantelis, Wood and McGorry2013; Hartmann et al., Reference Hartmann, Nelson, Spooner, Paul Amminger, Chanen, Davey, McHugh, Ratheesh, Treen, Yuen and McGorry2017). In other words, the diagnostic threshold for major depression does not align with stage 2 depression as conceptualised in clinical staging models (Hickie et al., Reference Hickie, Scott, Hermens, Naismith, Guastella, Kaur, Sidis, Whitwell, Glozier, Davenport, Pantelis, Wood and McGorry2013; Verduijn et al., Reference Verduijn, Milaneschi, van Hemert, Schoevers, Hickie, Penninx and Beekman2015; Hartmann et al., Reference Hartmann, Nelson, Spooner, Paul Amminger, Chanen, Davey, McHugh, Ratheesh, Treen, Yuen and McGorry2017). Nevertheless, it is of value to review the findings to date in order to inform future prediction research.
In FHR studies, a common identified clinical predictor of a first episode of MDD was the severity of subthreshold depressive symptoms (Warner et al., Reference Warner, Weissman, Fendrich, Wickramaratne and Moreau1992; Weissman et al., Reference Weissman, Fendrich, Warner and Wickramaratne1992; Colich et al., Reference Colich, Kircanski, Foland-Ross and Gotlib2015). Other identified clinical predictors included a ‘difficult’ temperament (Bruder-Costello et al., Reference Bruder-Costello, Warner, Talati, Nomura, Bruder and Weissman2007), irritability and anxiety (Rice et al., Reference Rice, Sellers, Hammerton, Eyre, Bevan-Jones, Thapar, Collishaw, Harold and Thapar2017) and behavioural (conduct) disorder (Warner et al., Reference Warner, Weissman, Fendrich, Wickramaratne and Moreau1992; Williamson et al., Reference Williamson, Birmaher, Axelson, Ryan and Dahl2004). Furthermore, low reward seeking (Rawal et al., Reference Rawal, Collishaw, Thapar and Rice2013), and broody rumination (Gibb et al., Reference Gibb, Grassia, Stone, Uhrlass and McGeary2012), were associated with the onset of depression.
Stressful or adverse life events were identified as a psychosocial predictor of onset of MDD in offspring at high FHR for depression (Hammen et al., Reference Hammen, Burge and Adrian1991; Rice et al., Reference Rice, Sellers, Hammerton, Eyre, Bevan-Jones, Thapar, Collishaw, Harold and Thapar2017). Furthermore, parents’ psychopathology such as severity of depression (Rice et al., Reference Rice, Sellers, Hammerton, Eyre, Bevan-Jones, Thapar, Collishaw, Harold and Thapar2017), an additional family member with depression (Rice et al., Reference Rice, Sellers, Hammerton, Eyre, Bevan-Jones, Thapar, Collishaw, Harold and Thapar2017) or presence of maternal anxiety disorder (Williamson et al., Reference Williamson, Birmaher, Axelson, Ryan and Dahl2004), has been found to predict new onset depression. Low socioeconomic status has also been associated with depression onset in these studies (Rice et al., Reference Rice, Sellers, Hammerton, Eyre, Bevan-Jones, Thapar, Collishaw, Harold and Thapar2017).
UHR studies: stage 1b
In contrast to psychosis prediction research, few at-risk studies of depression focused on the actual prodrome of depression and, when they have, they have largely involved adult populations (Fava and Tossani, Reference Fava and Tossani2007; Syed Sheriff et al., Reference Syed Sheriff, McGorry, Cotton and Yung2015). Although falling outside the scope of the current review, it should be noted that several trials have tested the effectiveness of preventative interventions in young people conceptually at UHR for depression. At-risk young people in these prevention trials have been mostly defined according to a parent's family history of depression (selective prevention, presumably stage 0) combined with the person's own history of depression or current depressive symptoms (indicated prevention, presumably at stage 1) (Clarke et al., Reference Clarke, Hornbrook, Lynch, Polen, Gale, Beardslee, O'Connor and Seeley2001; Garber et al., Reference Garber, Clarke, Weersing, Beardslee, Brent, Gladstone, DeBar, Lynch, D'Angelo, Hollon, Shamseddeen and Iyengar2009; Beardslee et al., Reference Beardslee, Brent, Weersing, Clarke, Porta, Hollon, Gladstone, Gallop, Lynch, Iyengar, DeBar and Garber2013). However, there appear to be no equivalent formal operationalisation of an ‘UHR’ state in depression as in psychosis or marginally in BD.
Other enriched cohorts: stage 1a/b
As subthreshold depressive symptoms are a strong and consistent predictor of subsequent MDD (Gotlib et al., Reference Gotlib, Lewinsohn and Seeley1995; Angst and Merikangas, Reference Angst and Merikangas1997; Pine et al., Reference Pine, Cohen, Cohen and Brook1999; Lewinsohn et al., Reference Lewinsohn, Solomon, Seeley and Zeiss2000; Cuijpers et al., Reference Cuijpers, Smit and Willemse2005; Fergusson et al., Reference Fergusson, Horwood, Ridder and Beautrais2005; Georgiades et al., Reference Georgiades, Lewinsohn, Monroe and Seeley2006; Keenan et al., Reference Keenan, Hipwell, Feng, Babinski, Hinze, Rischall and Henneberger2008; Johnson et al., Reference Johnson, Cohen and Kasen2009; Seeley et al., Reference Seeley, Stice and Rohde2009; Shankman et al., Reference Shankman, Lewinsohn, Klein, Small, Seeley and Altman2009), more recent at-risk approaches define depression risk by virtue of having subthreshold depressive symptoms (Klein et al., Reference Klein, Shankman, Lewinsohn and Seeley2009; Hill et al., Reference Hill, Pettit, Lewinsohn, Seeley and Klein2014). Two recent cohort studies prospectively investigated predictors of onset of MDD in at-risk groups defined by a combination of familial, subthreshold symptomatology and environmental risk factors (Goodyer et al., Reference Goodyer, Herbert, Tamplin and Altham2000a, Reference Goodyer, Herbert, Tamplin and Altham2000b; Carbonell et al., Reference Carbonell, Reinherz, Giaconia, Stashwick, Paradis and Beardslee2002; Goodyer et al., Reference Goodyer, Bacon, Ban, Croudace and Herbert2009; Goodyer et al., Reference Goodyer, Croudace, Dudbridge, Ban and Herbert2010). As with the FHR studies, their aim was to identify antecedents of the illness that could be used to target individuals for preventative intervention.
Identified antecedents/predictors of progression to stage 2 depression
The clinical and psychosocial predictors identified in these approaches are similar to the FHR studies. These include more severe depressive symptoms (Goodyer et al., Reference Goodyer, Herbert, Tamplin and Altham2000a, Reference Goodyer, Herbert, Tamplin and Altham2000b; Goodyer et al., Reference Goodyer, Croudace, Dudbridge, Ban and Herbert2010) and recent stressful life events (Goodyer et al., Reference Goodyer, Herbert, Tamplin and Altham2000b; Goodyer et al., Reference Goodyer, Croudace, Dudbridge, Ban and Herbert2010). In a study that selected at-risk adolescents based on subthreshold depressive symptoms, several subgroups were identified. In young people with poor social support, the highest risk of transition to full threshold depression was among those with a history of anxiety and substance use disorders. Among those with good social support, the highest risk was among those who reported multiple negative life events in the past year or a history of anxiety disorder (Hill et al., Reference Hill, Pettit, Lewinsohn, Seeley and Klein2014).
The predominant at-risk approach in depression is at stage 0, i.e. young people at familial high risk. Other enriched cohorts are based on the presence of subthreshold depressive symptoms; however, no formal operationalisation of an UHR approach has been identified. Main predictors of depression onset in at-risk samples include (in addition to a positive family history and subthreshold depressive symptoms) anxiety and stressful life events.
Discussion
Clinical staging models have gained substantial traction in recent psychiatric research, providing an alternative to the rigidity and shortcomings of traditional (DSM or ICD-based) diagnostic categories (McGorry, Reference McGorry2007; Hyman, Reference Hyman2010, Reference Hyman2011). However, an ongoing debate has focussed on whether clinical staging models should be general/transdiagnostic or disorder-specific. This debate has been somewhat hampered by the lack of cross-diagnostic discourse and empirical study design. The current review aimed to facilitate such cross-diagnostic discourse by scoping the body of literature of existing prospective at-risk approaches in young people and summarising identified clinical antecedents/predictors across diagnostic silos (psychosis, BD, depression) and placing these in the context of clinical staging.
It is evident from this scoping review that there has been extensive at-risk prediction research in all three disorder groups, with FHR studies (i.e. prediction from stages 0 to 2) present in all three disorders but predominantly in psychosis and BD. The clinical trajectory of those at familial high risk (stage 0) who progress to stage 2 disorder should be examined in future research from the perspective of the clinical staging model, i.e. do these cases progress through a period of general distress, help-seeking, functional decline and subthreshold symptoms (stages 1a and b) before developing stage 2 disorder and, if so, what are the opportunities for identification and intervention. Putting aside this issue of stage progression in FHR individuals, there are several limitations to the FHR approach in predicting psychosis and BD onset. First, the transition rate from non-enriched samples of familial high-risk youth to full-threshold BD or psychosis is relatively low (9.6–11%) (Mesman et al., Reference Mesman, Nolen, Reichart, Wals and Hillegers2013; Duffy et al., Reference Duffy, Horrocks, Doucette, Keown-Stoneman, McCloskey and Grof2014; Axelson et al., Reference Axelson, Goldstein, Goldstein, Monk, Yu, Hickey, Sakolsky, Diler, Hafeman, Merranko, Iyengar, Brent, Kupfer and Birmaher2015). Second, the prevalence of a first-degree family history of BD or psychosis among clinical samples is low (Ratheesh et al., Reference Ratheesh, Cotton, Davey, Adams, Bechdolf, Macneil, Berk and McGorry2017a) indicating that using the FHR approach alone may miss a large proportion of those who ultimately develop BD or psychosis.
While UHR research (i.e. prediction from stages 1b to 2) is a widely adopted approach in psychosis prediction, there have been fewer systematic approaches to define an ‘ultra’-high (i.e. explicitly aiming to capture the putative prodrome of the disorder) risk concept in BD, and there is virtually no standardised, operationalised approach in MDD. This may partly reflect the fact that early intervention efforts (and the requisite case identification approaches) initially focused on early psychosis and have only evolved to include non-psychotic disorders (within a broader youth mental health framework) in more recent years (McGorry et al., Reference McGorry, Goldstone, Parker, Rickwood and Hickie2014b). However, increasing focus is now being applied to defining ‘UHR’ for mood disorders, particularly BD (Ratheesh et al., Reference Ratheesh, Davey, Hetrick, Alvarez-Jimenez, Voutier, Bechdolf, McGorry, Scott, Berk and Cotton2017b).
The review identified a range of clinical risk factors of disease progression. Some of these antecedents were specific to particular disorders, whereas some were shared. Among the specific predictors were subthreshold manic symptoms for BD and attenuated psychotic symptoms (particularly in the domains of conceptual disorganisation and unusual thought content) for psychosis, thus representing subthreshold versions of the respective disorder. It has been argued that these sub-threshold symptoms crystallise over time into full syndromal disorder and may represent later stage phenotypes on the at-risk continuum (Yung et al., Reference Yung, Yuen, McGorry, Phillips, Kelly, Dell'Olio, Francey, Cosgrave, Killackey, Stanford, Godfrey and Buckby2005; McGorry, Reference McGorry2014).
Among the common identified antecedents were anxiety, depression, sleep disturbance, poor psychosocial functioning, substance misuse, trauma/stressful life events and having an affected family member (inherent to selected study designs). Although the specificity for particular stage 2 outcomes yielded by these shared variables may be modest, their sensitivity for identifying people at risk of stage 2 disorder generally appears to be high. These early overlapping, non-specific antecedents may provide support for a broad-based, transdiagnostic clinical staging model, at least in the early stages of serious psychotic and mood disorders (stages 0–1). Indeed, an increasing body of literature (not reviewed here) defines outcome as well as at-risk mental states more broadly using various approaches such as simultaneously including offspring of parents affected by multiple disorders (e.g. Sanchez-Gistau et al., Reference Sanchez-Gistau, Romero, Moreno, de la Serna, Baeza, Sugranyes, Moreno, Sanchez-Gutierrez, Rodriguez-Toscano and Castro-Fornieles2015; Paccalet et al., Reference Paccalet, Gilbert, Berthelot, Marquet, Jomphe, Lussier, Bouchard, Cliche, Gingras and Maziade2016), mainly informed by findings of molecular-genetic overlap between disorders (Moskvina et al., Reference Moskvina, Craddock, Holmans, Nikolov, Pahwa, Green, Owen and O'Donovan2009; Pettersson et al., Reference Pettersson, Larsson and Lichtenstein2016).
A formal operationalisation of a broad, pluripotent at-risk mental state (Hartmann et al., Reference Hartmann, Nelson, Spooner, Paul Amminger, Chanen, Davey, McHugh, Ratheesh, Treen, Yuen and McGorry2017) has the advantage of identifying help-seeking young people with a range of sub-threshold clinical presentations and allowing for heterotypic continuity [one sub-threshold disorder evolving into divergent stage 2 outcomes, which is commonly observed (Lin et al., Reference Lin, Wood, Nelson, Beavan, McGorry and Yung2015; Rutigliano et al., Reference Rutigliano, Valmaggia, Landi, Frascarelli, Cappucciati, Sear, Rocchetti, De Micheli, Jones, Palombini, McGuire and Fusar-Poli2016)]. Furthermore, focusing on a broad range of disorders as outcome of interest addresses the problem of low incidence rates (and hence low statistical power) when focusing on single outcome such as schizophrenia (Cuijpers, Reference Cuijpers2003). As such, it addresses some of the recent concerns expressed about the ‘UHR’ for psychosis paradigm – that is, the modest psychosis transition rates and that this identification approach places undue emphasis on psychotic disorder as an outcome of attenuated psychotic symptoms in the context of a range of a multidimensional psychopathology (e.g. van Os and Guloksuz, Reference van Os and Guloksuz2017).
A transdiagnostic at-risk approach is certainly not without disadvantages and risks. First of all, more research is required addressing the question whether there is sufficient ground for pluripotency of early clinical phenotypes (Duffy and Malhi, Reference Duffy and Malhi2017). Furthermore, it has been argued that broad approaches are not well suited to accommodate intervening variables such as substance misuse (Scott and Henry, Reference Scott and Henry2017) and that ‘lumping’ of heterogeneous disorders may mask disorder-specific findings while there is not enough evidence for pluripotency (Duffy and Malhi, Reference Duffy and Malhi2017). However, broad, inclusive models do not preclude examination of specific sub-type trajectories as long as research studies taking this approach remain sufficiently fine-grained in their assessments to pick up specific signals amongst noise (Duffy et al., Reference Duffy, Malhi and Grof2017). While certain characteristics (e.g. lithium responsiveness) are likely to confer specificity in prevention paradigms, these may be studied as sub-streams within the broader, pragmatic pluripotent at-risk approaches.
Recently, the view that psychopathology can be seen as networks of symptoms dynamically interacting and influencing each other, rather than being the expression of an underlying latent ‘cause’ (i.e. the mental disorder), has gained traction (Borsboom and Cramer, Reference Borsboom and Cramer2013; Borsboom, Reference Borsboom2017). This view is particularly interesting in the current context of the possible pluripotency of early psychopathology. How symptoms dynamically interact with and impact each other may depend on the stage of disorder, with weak, diffuse symptom-interactions occurring in the earlier stages, becoming stronger and more person-specific over time, eventually giving rise to diagnostic specificity in later stages (Wigman et al., Reference Wigman, van Os, Thiery, Derom, Collip, Jacobs and Wichers2013). These novel models of conceptualising psychopathology, along with other recent frameworks, such as RDoC and the Hierarchical Taxonomy of Psychopathology (Kotov et al., Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby, Brown, Carpenter, Caspi, Clark, Eaton, Forbes, Forbush, Goldberg, Hasin, Hyman, Ivanova, Lynam, Markon, Miller, Moffitt, Morey, Mullins-Sweatt, Ormel, Patrick, Regier, Rescorla, Ruggero, Samuel, Sellbom, Simms, Skodol, Slade, South, Tackett, Waldman, Waszczuk, Widiger, Wright and Zimmerman2017), rely on the use of transdiagnostic, dimensional and data-driven approaches that assess a range of psychopathology with continuous measures over extended periods of time. These approaches, which do not merely reflect and reproduce existing DSM nosology, may shed further light on the ‘lumping v. splitting’ issue raised here.
The current review of the literature was limited to clinical and psychosocial variables that characterise stages prior to stage 2 disorder and did not include other variables, such as neurocognitive, neurobiological or genetic variables. At this stage in the development of early intervention services, at-risk identification approaches are rooted in clinical ‘bedside’ assessments. Therefore, collating information on clinical predictors of stage 2 disorder across diagnoses is most useful in terms of clinical translatability over the short term. As a next step, the view can be extended to biological domains. The integration of different levels and domains of data, combining genetics, imaging, cognitive and clinical science, is the ultimate goal and has already shown some promise (Nieman et al., Reference Nieman, Ruhrmann, Dragt, Soen, van Tricht, Koelman, Bour, Velthorst, Becker, Weiser, Linszen and de Haan2014; Clark et al., Reference Clark, Baune, Schubert, Lavoie, Smesny, Rice, Schafer, Benninger, Feucht, Klier, McGorry and Amminger2016). This integrated approach, with the aim of producing a tool that can be implemented in clinical practice, is the focus of a number of consortia-based studies, such as EU-GEI (http://www.eu-gei.eu), PSYSCAN (https://www.pronia.eu/the-project) and PRONIA (http://www.psyscan.eu).
The current overview did not include prospective studies in unselected community samples (e.g. Kovacs et al., Reference Kovacs, Feinberg, Crouse-Novak, Paulauskas, Pollock and Finkelstein1984; Lewinsohn et al., Reference Lewinsohn, Hoberman and Rosenbaum1988; Reinherz et al., Reference Reinherz, Giaconia, Pakiz, Silverman, Frost and Lefkowitz1993; Orvaschel et al., Reference Orvaschel, Lewinsohn and Seeley1995; Hafner et al., Reference Hafner, Maurer, Loffler, an der Heiden, Munk-Jorgensen, Hambrecht and Riecher-Rossler1998; Oldehinkel et al., Reference Oldehinkel, Wittchen and Schuster1999; Reinherz et al., Reference Reinherz, Giaconia, Hauf, Wasserman and Paradis2000; Kim-Cohen et al., Reference Kim-Cohen, Caspi, Moffitt, Harrington, Milne and Poulton2003; Beesdo et al., Reference Beesdo, Hofler, Leibenluft, Lieb, Bauer and Pfennig2009; Zimmermann et al., Reference Zimmermann, Bruckl, Nocon, Pfister, Lieb, Wittchen, Holsboer and Angst2009; Tijssen et al., Reference Tijssen, van Os, Wittchen, Lieb, Beesdo, Mengelers and Wichers2010) or studies that only assessed the course of the condition or factors affecting symptom severity rather than new incidences of disorder. Although these studies may be valuable in identifying predictors of onset of mental illness, they fell outside the scope of the present paper. Furthermore, this work reviewed existing at-risk research approaches per disorder alongside antecedents of homotypic progression and therefore disregarded findings of integrated, transdiagnostic at-risk approaches or approaches investigating heterotypic development. Similarly, we were not able to address the body of literature on overlap in genetic risk (Moskvina et al., Reference Moskvina, Craddock, Holmans, Nikolov, Pahwa, Green, Owen and O'Donovan2009; Pettersson et al., Reference Pettersson, Larsson and Lichtenstein2016). While these approaches go to the heart of a general staging model, at least in the early stages of severe psychotic or mood disorders, they represent the minority of the research conducted to date and may be targeted in future reviews. Finally, we reviewed at-risk literature and predictors of BP, MDD and psychosis with the respective disorder as the single outcome of interest. Heterotypy of at-risk states (e.g. offspring of parents with psychotic illness who go on to develop depression) have not been considered. While the potential for heterotypy/multifinality is a central principle of a pluripotent at-risk mental state (Hartmann et al., Reference Hartmann, Nelson, Spooner, Paul Amminger, Chanen, Davey, McHugh, Ratheesh, Treen, Yuen and McGorry2017), this was outside the scope of this review and will be considered in future work.
In conclusion, this scoping review provided a broad overview of employed at-risk approaches and identified antecedents of illness progression in three major mental disorders (psychosis, BD and depression) in the context of clinical staging. Stage 0 at-risk conceptualisations (i.e. FHR approaches) were identified in all three disorders. However, formalised stage 1b conceptualisations (i.e. UHR approaches) were only present in psychosis and marginally in BD. The presence of non-specific and overlapping antecedents in the three disorders supports the development of a broad, pluripotent at-risk mental state rooted in the clinical staging model of psychiatry.
Financial support
J.A.H. was supported by a Netherlands Organization for Scientific Research (NWO)-Rubicon Grant (825.15.015). P.D.M. was supported by a Senior Principal Research Fellowship from the NHMRC (ID: 1060996) and B.N. was supported by a University of Melbourne Faculty Fellowship. P.D.M. reported receiving grant funding from the National Alliance for Research on Schizophrenia and Depression and unrestricted research funding from AstraZeneca, Eli Lilly, Janssen-Cilag, Pfizer and Novartis, as well as honoraria for educational activities with AstraZeneca, Eli Lilly, Janssen-Cilag, Pfizer, Bristol-Myers Squibb, Roche and the Lundbeck Institute.
Conflict of interest
None.
