Anticipating psychosis

Introduction

During the 20th century, the concept of schizophrenia gave rise to a plethora of hypotheses and research programmes, involving a wide range of disciplines and approaches. While the 1960s marked the culmination of the longstanding dispute over the relative emphasis to be placed on the psychology versus the biology of the disorder, explanations based exclusively on brain dysfunction and genetics dominated schizophrenia research by the early 1980s. Particularly pronounced in the USA, this swing was facilitated by the significant narrowing of the schizophrenia category in the neo-Kraepelinian DSM-III (Andreasen, 1989). The closing years of the century, however, showed a shift towards an understanding of schizophrenia as a complex phenotype with both biological and psychosocial determinants. In Europe especially, there has been a retreat from the idea of a distinct brain disease, and a clamour has broken out for a more dimensionally driven model of psychosis as a transdiagnostic spectrum that extends into the general population (Guloksuz and Van Os, 2017). In an effort to ameliorate patient outcomes, psychiatrists have furthermore turned their attention to risk estimation and early intervention in young people with mild psychotic symptoms. Yet, we are a long way from the vision of a personalized and pre-emptive ‘precision medicine’ modelled on advances in early cancer identification via molecular markers (Insel, 2013). Despite definitional changes and vigorous research efforts, progress in aetiological and therapeutic knowledge has been meagre, and disagreement persists on what constitutes the core features and limits of schizophrenia. Indeed, an increasing number of investigators now argue that the construct should be abandoned altogether.

In recent years, much controversy has centred on the burgeoning field of research on ‘prodromal’ or ‘ultra-high-risk’ states, and its translation into clinical practice. When it was proposed that a psychosis risk syndrome should be included in DSM-5, leading psychiatrists warned that this could lead to a public health catastrophe, given the demonstrated high false-positive rates and consequent danger of widespread stigmatization and overtreatment of young people with harmful drugs (Frances, 2012). Moreover, opponents of the influential early intervention movement have accused policy makers of uncritically accepting exaggerated claims about the effectiveness of specialized programmes for prodromal and first-episode patients, to the detriment of generic mental health services (Malla and Pelosi, 2010).

In the context of wider social and political trends, early intervention in psychosis stands out as an example of the rationales and practices developing around the themes of susceptibility, risk, and precaution described by Nikolas Rose (2010) under the heading of ‘screen and intervene’. Whereas early intervention advocates highlight the possibility of reducing human and social costs in the long term (McGorry, 2015), Rose and others have explored the potentially sinister implications of this emerging configuration, suggesting that future screening methods based on genetic profiling and brain scanning may result in threshold-lowering and the pre-emptive exclusion of undesirable individuals.

An observer of the debates surrounding early intervention might get the impression that the anticipation and timely treatment of mental illness is a relatively new priority. In fact, throughout the 19th century, alienists argued that delay in seeking help could be disastrous (Scull, 1979: 111–12). In the first half of the 20th century, salvation of the potentially deranged became the objective of the mental hygiene movement, and psychiatrists discussed the prospect of identifying subjects in the prodromal stage (Cameron, 1938; Mayer-Gross, 1932; Sullivan, 1927). Although this is sometimes acknowledged, it is rarely recognized that the current clinical paradigm was foreshadowed by an ambitious high-risk research programme which drew support from a number of major funding bodies. Looking primarily at children of women diagnosed with schizophrenia, high-risk researchers hoped to identify precursors and causal risk factors that could ultimately inform the development of effective prevention strategies. As it turned out, their studies signalled the emergence of a durable psychopathological fringe category but produced few robust results on which to base precise prognosis and intervention.

This article analyses the rise and decline of the high-risk research movement, focusing on the innovative Copenhagen High-Risk Project started in 1962 by American psychologist Sarnoff Mednick (1928–2015) and Danish psychiatrist Fini Schulsinger (1923–2012). The first half of the article examines the background of the project, notably the methodological issues it addressed, and the way it relied on the administrative infrastructure of the Danish welfare state. Then follows a discussion of the study’s approach and findings, which inspired an array of investigators with diverse agendas to set up their own prospective risk studies. The final section examines the context of the eventual termination of the high-risk research programme towards the late 1980s, including the immense difficulties associated with the longitudinal design; the frequent failures to replicate findings across studies; and the uncertainties and ambiguities inherent in the concepts of ‘vulnerability’ and ‘high risk for schizophrenia’.

The Copenhagen High-Risk Project: Background and key figures

Having explored the nature of creativity as a young researcher at UC Berkeley, Sarnoff Mednick became increasingly known during the 1960s for his theories on the genesis of schizophrenia. Building on insights from learning theory and experimental studies of arousal and conditioning in adult schizophrenics, Mednick (1958) conceptualized chronic schizophrenia as the result of a learned pattern of avoidant thinking that enabled the vulnerable individual to cope with intense fearfulness in the face of stress. He then began to question the prevalent research methodology which largely included three approaches: (a) studying already diagnosed, hospitalized schizophrenics and comparing them with normal controls; (b) examining ‘schizophrenogenic’ patterns of communication in the families of schizophrenics, inferred from psychoanalytic theories; and (c) analysing the records and life histories of schizophrenics retrospectively. Rather than yielding clues on aetiology, he noted, the first two confounded cause and effect and revealed more about the consequences of illness, that is, social failure, family dysfunction, medication side effects, poor physical health and long-term institutionalization (Mednick, 1968). Indeed, contrary to the then widespread practice of highlighting the mother’s pathological influence, he emphasized that parents might exhibit deviant behaviour because of the sick child.

The proposed solution was to study individuals prior to the onset of manifest disorder and follow them over time – a strategy that also bypassed the drawbacks of retrospective research, as it allowed for the collection of many kinds of data and abolished the need to rely on the fallible memories of patients and their relatives. Significantly, given the low base rate of schizophrenia in the general population (1–2%), Mednick suggested that such demanding research concentrate on vulnerable ‘high-risk’ groups with schizophrenic parents. This idea was partly inspired by the literature on psychiatric genetics – notably the family research of Franz Kallmann (1938), which at the time was still much maligned among American psychiatrists and psychologists due to its association with Nazi eugenic politics – as well as a paper by Pearson and Kley (1957) asserting that the propensity of abnormal behaviour to run in families should be exploited in longitudinal studies. His intention was to uncover a pattern of premorbid characteristics that reliably identified the children who fell ill, and on this basis to suggest a theory of aetiology which, in turn, meant ‘that our interventions would not be random, do-good, unspecific, hit-or-miss programs’ (Mednick, 1971: 50).

Several influences intersected in Mednick’s research proposal, including the growing interest in genetic contributions to major mental illness vis-a-vis the environment, and the contemporary emphasis within American psychology on the experimental study of schizophrenia which emerged with David Shakow and associates’ research programme at Worcester State Hospital during the 1940s (Cautin, 2008). Though the risk strategy was conceived in opposition to existing methodologies, it moreover echoed the longstanding American preoccupation with premorbid conditions that had characterized elite discourses on mental disorder since Adolf Meyer (Noll, 2017: 339). Indeed, compared with their European counterparts, American psychiatrists throughout most of the 20th century retained a curious lack of concern with detailing the ongoing signs, symptoms and courses of psychosis after its onset. Still, Mednick’s approach was quite novel. Other investigators had examined the consequences for children of having mentally ill parents, but the systematic study of children of schizophrenics was a neglected area of research, and his plan to track a large group of subjects and predict who would break down was perceived as unusual and bold, if not overambitious.¹

That is not to say that Mednick was the first to focus on children with familial risk of schizophrenia. In 1941, Manfred Bleuler started a major follow-up study of 208 of his schizophrenic patients, in which he included their children (184 altogether) (Bleuler, 1984). The study, replete with Bleuler’s personal and often moving observations, investigated the effects of living with a schizophrenic parent, and differed from all previous research in finding a much higher proportion (75%) of healthy, full-grown offspring. Ten years later in New York, child psychiatrist Barbara Fish set out to test her hypothesis that the developmental abnormalities she detected in a group of infants born to schizophrenic mothers were related to an inherited neuro-integrative defect that caused susceptibility to the illness (Fish, 1957). By contrast, David Sobel’s (1961) study of infants with two schizophrenic parents – half reared by their biological parents and half in foster homes – emphasized interaction with those parents as the most likely mechanism for the transmission of disorder. But whereas these studies dealt with small numbers of subjects or/and lacked independent observers and appropriate control groups, Mednick invested heavily in more objective methods which yielded quantitative data. Fish also had problems obtaining funding for her study, partly due to her claim that schizophrenia could be predicted from the behaviour and functioning of 1-month-old babies – a proposition that was difficult to accept for critics who struggled to believe that factors so disconnected in time could be closely related. Starting with adolescent subjects, Mednick had more success a few years later after encountering significant obstacles.

With colleague Jerry Higgins, Mednick first tried to launch his project in Detroit but quickly became frustrated. The records in Michigan were inadequate for the purpose, and due to the high residential mobility rate across cities and states in the USA it was impossible to follow participants for an extended period of time (Mednick and Schulsinger, 1974: 52). The need to get separate permissions from various authorities also created difficulties. But then an address by the renowned Danish psychiatric geneticist, Erik Strömgren, directed Mednick’s attention to the possibilities afforded by the national population and disease registers in Scandinavia. ² In 1961, he went to Copenhagen to present his study and here met the new young Head of Psychiatry at the Municipal Hospital (Kommunehospitalet), Fini Schulsinger. Schulsinger quickly saw the potential, and steps were taken to locate suitable subjects and secure financial support. For the first year’s work alone, they calculated with expenses of 45,000–60,000 US dollars, corresponding to about half the total amount allocated for medical research by the Danish Science Foundation (Videnskabsfonden) at this time ( Berlingske Tidende, 1962). The funds were provided by the National Association for Mental Health, the Scottish Rite Committee for Research on Schizophrenia, The Human Ecology Fund and the National Institutes of Mental Health. Yet, many of Mednick’s American colleagues had been sceptical of the endeavour from the very beginning. Some questioned his speculative, global theorizing on aetiology and others highlighted the logistical and scientific challenges involved (Garmezy in Mednick, 1960: 74–5). At this point, Mednick was already acquiring a reputation as an adventurer who would take big risks once excited by a project (Watt, 2014).

Schulsinger, who was appointed Secretary General of the World Psychiatric Association in 1983, stood out from the rest of the Danish psychiatric profession due to his informal manner, vocal political engagement, high research activity, and international outlook. Like Mednick, he came from a modest Jewish immigrant background. He had strong socialist sympathies as a young man and participated in the resistance movement during the war, first in Copenhagen and later in Stockholm where he fled in 1943 (Schulsinger H., 2014). Early in his career, he travelled to the USA on a WHO scholarship and was inspired by the dynamic atmosphere and responsiveness to innovative ideas there (Parnas, 2014). But while impressed with the potential of the new biochemical, physiological, and psychological avenues of schizophrenia research, Schulsinger retained a focus on the influence of social factors. He regularly called attention to the deficiencies of the Danish welfare system, for example when protesting about the overcrowded conditions in Copenhagen’s acute psychiatric wards, and moreover lamented the meagre funding for psychiatric research in Denmark: ‘When one includes the loss of productivity and the human suffering caused by this illness, it is strange that society does not actually PUSH physicians, psychologists, and other scientists to commit to researching schizophrenia’ (Schulsinger, 1968). Schulsinger’s main strengths were not those of the classical scholar but those of the clinician whose powers of persuasion and organizational skills could open doors and move things along. While engaged in various studies, he continued to run the hospital psychiatric ward and remained involved in public debates, criticizing the neglect of marginalized groups and defending the opportune Danish conditions for register-based research against those who argued that access should be restricted due to privacy concerns (Schulsinger, 1981).

In the collaboration with Mednick, Schulsinger played the more pragmatic part, but he, too, was an entrepreneurial type who did not dwell much on the problems that might arise (Higgins, 2015; Venables, 2014). In 1962, they established the independent Psykologisk Institut as a framework for what would soon become a highly dynamic, international research environment. Around the same time, Mednick arranged a meeting between Schulsinger and Seymour Kety, the influential NIMH Chief of Clinical Science, who proceeded to plan another famous collaboration, with David Rosenthal and Paul Wender: the Danish Adoption Studies of the relatives and adopted-away offspring of schizophrenics. Attempting what the high-risk project could not, the aim of these studies was to separate the effects of heredity and environment. Apart from generating acceptance of the role of genetic factors in schizophrenia and fostering similar research in Denmark on other forms of deviant behaviour, ³ they introduced the concept of a ‘genetic spectrum’ of psychopathology to capture the subtle signs identified in relatives. Thus, both the high-risk and adoption projects contributed to the broadening of focus within research, from hospitalized chronic patients to family members and prodromal subjects, as well as individuals in the general population with psychosis-like experiences or schizotypic characteristics. The latter would identified via a growing number of psychosis proneness rating scales, such as those developed by Loren and Jean Chapman in the USA, and by Gordon Claridge and colleagues in the UK.

An epidemiologist’s dream

Boasting a strong research profile in genetics and psychiatric epidemiology, Denmark offered many advantages as a site for the research. Regularly described as ‘an epidemiologist’s dream’ in international science journals (Bauer, 2014; Frank, 2000), the country had a homogenous population of about 5 million inhabitants, with a low rate of mobility and emigration. It also combined a universal welfare system with a long history of keeping detailed records of demographic and health-related events. Key sources of information in the high-risk and adoption studies were the Central Psychiatric Register of hospital admissions and the Central Population Register which kept track of changes of name and address. ⁴ The psychiatric register emerged as part of the private eugenic society, Den Antropologiske Komité, founded by physician Søren Hansen in 1904. From 1918, the society started registering individuals with mental retardation, serious mental disturbance, and defects such as deafness and epilepsy. In 1938, the Institute of Human Genetics [Institut for Human Arvebiologi og Eugenik] was established at the University of Copenhagen with funding from the Rockefeller Foundation. Here, under the leadership of geneticist Tage Kemp, the disease register grew and came to form the basis of a comprehensive genetic counselling service associated with the administration of the Pregnancy Measures Act of 1937 and the Sterilization Acts of 1934 and 1935. Providing access to abortion on medical and eugenic grounds and allowing sterilization of feebleminded individuals, this legislation was considered an important element in the major social welfare reform that was introduced in 1930s. ⁵ The registration of all state psychiatric hospital admissions began at the institute in 1953, and the records were later moved to the Department of Psychiatric Demography, led by Professor Strömgren, at Aarhus Psychiatric Hospital.

With the advent of electronic data processing, the Central Population Register was computerized in 1968 through the introduction of personal identification (CPR) numbers. Originally created to facilitate the deduction of income taxes at the source, CPR numbers were increasingly employed in research to track individuals, groups, and even large populations, and to examine aspects of their medical or social history through the linkage of data from a growing number of electronic registers. ⁶ From the beginning, this infrastructure attracted a numerous foreign investigators, notably from US agencies, who wished to exploit the favourable access conditions and high participation rates in research in Denmark.

From the epidemiologist’s perspective, a key attraction was the high level of trust that characterized Danish society, and the public’s notable willingness to contribute personal information to population health investigations. Schulsinger (1980, 1981) described this willingness as a reflection of the ‘mutual insurance principle’ – an informal contract between the individual and the welfare state – and furthermore insisted that gene–environment studies like those conducted at Psykologisk Institut were ethically justified when used to facilitate supportive measures targeted at vulnerable groups. These comments appeared in the 1980s in the context of a heated public debate about registers and register-based research, sparked by rising concerns over data protection and surveillance in the new electronic age. Nevertheless, Danish attitudes towards privacy rights and the possibility of data leakage or misuse have remained relatively relaxed, despite the intensifying sourcing and use of health data for research purposes in recent decades (Høyer, 2016).

Outline of the study

Even before the researchers started contacting the selected families, the high-risk study was widely described in the Danish media. With barely disguised national pride, reporters conveyed Mednick’s enthusiasm for the project and the Danish resources, adding that high-tech equipment otherwise used to measure the physiological reactions of American astronauts – the polygraph – had been flown in from the USA ( Berlingske Aftenavis, 1962; Berlingske Tidende, 1962). All features highlighted the intriguing prospect of stopping schizophrenia in its tracks at an early age.

The study sample included 207 children with an average age of 15, born to mothers diagnosed with ‘chronic’ schizophrenia and located through the Central Psychiatric Register. ⁷ These cases were matched with 104 low-risk controls with similar backgrounds but no known mental illness in the immediate family. From the works of Kallmann and a preliminary test study in Denmark (Reisby, 1967), it was inferred that about 15% of the high-risk group would come to suffer from schizophrenia themselves and that half would eventually manifest some psychiatric disturbance. While requiring biological supports, allegiance to a genetic model was not required for Mednick’s hypothesis of heightened autonomic responsivity. However, the genetic rationale became clearer with time, especially as the results from the adoption studies started coming in.

The first round of assessments in 1962 comprised psychophysiological skin conductance measures of responses to unpleasant noise stimuli; a battery of psychological tests (intelligence, personality, and word association); psychiatric evaluations; and social worker interviews with parents and guardians (Mednick, 1968). Life histories, school behaviour reports, blood samples, and midwife reports were also collected. Casting the net wide, the researchers took full advantage of the Danish welfare system and gathered all the data that the literature suggested might be relevant (Mirdal, 2014). The risk element was not explained to the subjects or their families, unless they asked directly (Mednick, 2014). As discussed in a NIMH report on the ethical implications of high-risk research, this issue presented a vexing moral dilemma (Curran, 1973: 80–2). There was a reluctance to indicate that the vulnerability might be genetic in origin; when recruited, parents were simply told that the study looked at the effects of ‘nervous breakdown’ on family members (Mednick and Schulsinger, 1974: 56).

In the psychophysiological tests, high-risk children were found to react more strongly to stress stimuli, but the consistent finding that their skin conductance responses also recovered faster than those of controls completely contradicted Mednick’s original thesis, which therefore had to be revised. All subjects were reassessed in 1967, at which point 20 high-risk subjects, referred to as the ‘sick group’, had suffered some form of mental breakdown. Overall, the sick group tended to have lost their schizophrenic mothers to hospitalization earlier in life; to have displayed disruptive behaviour in class (a pattern that did not quite fit the old stereotype of the shy and ‘shut-in’ pre-schizophrenic child personality); and to have ‘drifted’ markedly on the continuous association test. On electrodermal measures, the majority showed rapid response to stimulation, and no sign of habituation. However, the sick offspring were distinguished most strikingly of all from high-risk subjects who had remained well by a much higher rate of severe birth complications. This was unexpected, as there was little in the schizophrenia literature to suggest such a link. Research assistant Gretty Mirdal had previously studied Benjamin Pasamanick and colleagues’ ‘continuum of reproductive causality’ thesis, and it was in fact she who persuaded Mednick and Schulsinger that the midwife reports warranted a closer analysis (Mednick, 2014). Offering a real possibility of establishing effective prevention programmes focused on obstetric care, the finding generated much excitement and publicity (Mednick, 1971), even if the initial enthusiasm was tempered by studies that identified no significant differences in the rates of perinatal complications between the offspring of psychotic parents and normal controls (Mirdal et al., 1977), and between women with schizophrenia and those with different psychiatric disorders (Zax et al., 1977).

From playing no role in the original theory, perinatal complications now came to occupy a central position as Mednick put forward another provocative hypothesis. Based on animal models, he suggested that perinatal anoxia might damage the hippocampus and thus the modulatory control of the body’s stress response in genetically vulnerable individuals. Some years later, neurological insult from flu infections during the second trimester of pregnancy was highlighted as a likely additional risk factor (Mednick et al., 1988). These findings eventually led to new formulations assuming an early disruption in normal brain development, secondary to a combination of genetic and environmental factors. In subjects with the most severe outcomes interactions, pathogenesis seemed to involve genetic predisposition, often involving psychopathology in the father also; ⁸ pregnancy and birth complications; and early parental separation/institutional rearing. In the 1980s, when new brain imaging techniques became available, computed tomography assessments of the high-risk subjects revealed a higher rate of ventricular enlargement in the schizophrenia group, notably in those who had suffered delivery complications (Cannon and Mednick, 1993). Significantly, this material heavily inspired the influential neurodevelopmental hypothesis of schizophrenia, sometimes caricaturized as the ‘doom from the womb’ hypothesis (Murray, 2015, Murray and Lewis, 1987).

From 1972 through 1974, psychologist Hanne Schulsinger (1976) – Fini Schulsinger’s wife – performed a series of interviews with the subjects, using three different diagnostic instruments. ⁹ Consensus diagnoses identified 15 high-risk individuals as schizophrenic and 55 as displaying ‘borderline’ or ‘latent’ states (corresponding numbers for the normal controls were one and four respectively). Seven in the high-risk group had committed suicide by the end of the reassessment period, while none of the low-risk subjects had died. Employing the stricter DSM-III criteria at the 25-year follow-up in 1986–1989, 31 high-risk subjects were diagnosed with schizophrenia and 36 with the milder schizotypal personality disorder (Parnas et al., 1993). Interestingly, among the most powerful predictors were teacher reports showing that pre-schizophrenic males were more likely to have been disruptive and emotionally excitable in class, whereas female pre-schizophrenics were judged to be abnormally nervous, passive, and socially withdrawn (Olin and Mednick, 1996). The demonstration of such continuity in social dysfunction and behavioural deviance lent additional support to the conceptualization of schizophrenia as a gradual development, rather than an adult-onset degenerative disease that suddenly strikes the healthy person.

The emergence of the high-risk research movement

The first results of the Copenhagen study were presented in 1967 in Dorado Beach, Puerto Rico, at a NIMH-supported confrontation conference where a group of leading investigators with disparate disciplinary backgrounds were invited to present their data and defend their positions on the transmission schizophrenia (Rosenthal and Kety, 1968). The heated discussions that took place helped make sure that the event quickly achieved a myth-like status. Like the conference on the origins of schizophrenia convened by John Romano in Rochester earlier the same year, it attempted to synthesize biological and psychosocial perspectives and reflected the divergent lines of intramural research at the NIMH during this period.

In 1972, another landmark conference was held in Dorado, this time bringing together key investigators in the nascent high-risk research field. Its main facilitator was Loren Mosher (1933–2004), the director of the newly formed extramural Center for Studies of Schizophrenia, who later became (in)famous for this attempt to create and validate a therapeutic facility for schizophrenic patients modelled after R. D. Laing’s Kingsley Hall. While he admired Laing and found the biological position lacking, Mosher was highly interested in high-risk research and played an important role in obtaining support from the NIMH (Mosher and Wynne, 1970). In addition to organizing the first conferences on the subject, he fostered a lengthy authoritative review prepared by Norman Garmezy (1918–2009) (Garmezy, 1974), who mentored numerous researchers in developmental psychology. Garmezy was, perhaps more than any other person, responsible for advocating the value of prospective research. In 1973, with the participation of the William T. Grant Foundation, he headed the creation of a Risk Research Consortium that would encourage coordination between projects and organize informal meetings on methodological and practical issues. In 1972, he started his own project in Minnesota, the results of which came to form the basis of his well-known research programme on competence and resilience in children.

Thus, inspired by the promising results of the Copenhagen study, developmental-prospective high-risk research grew into a movement backed by the NIMH and other prominent funding bodies. By the end of 1973, more than 20 different studies had been launched in the USA and a few European countries, most of them on children of schizophrenia patients. The high-risk strategy attracted many of the post-war period’s most talented psychologists, and expectations were high that it would facilitate a leap forward in the understanding of schizophrenia. Taking their cue from existing experimental studies conducted with adult patients, researchers were interested in measuring traits that could be linked to full-blown schizophrenia, notably attention and information-processing impairments, and signs of cognitive slippage. Nonetheless, studies varied greatly in terms of theoretical orientation, sample characteristics, and variables. The only study that did not base the evaluation of risk on schizophrenic parentage was the UCLA Family Project, which emphasized distorted communication in the families of adolescents seen in a clinic (Rodnick et al., 1984). At the other end of the spectrum was the New York High-Risk Project which tried to identify attentional-cognitive, neuro-motor and psychophysiological dysfunction markers thought to reflect underlying genetic defects (Erlenmeyer-Kimling and Cornblatt, 1987). Disagreements were common among consortium members, but at this point, before the mapping of the genome became a key objective and the 1990s were heralded ‘the decade of the brain’, there was considerable support for such theoretical diversity and debate at the NIMH (Watt, 2014).

To fully understand the attraction of Mednick and Schulsinger’s work, one needs to look beyond the already acknowledged problems of studying already diagnosed individuals. Apart from including large samples and a wide range of measures, the Copenhagen project built on an aetiological theory which seemed to permit validation, and it identified variables that lent themselves to immediate reliability testing. It turned up several remarkable findings, and Mednick’s use of the evocative (and provocative) term ‘high risk’ gave this type of research a name and a sense of identity. Resonating with the NIMH’s stated goal to further interdisciplinary research, the elasticity of the concept at the same time made it possible to test divergent hypotheses within the same conceptual framework. More generally, the rise of prospective-longitudinal research on children at risk for schizophrenia reflected a revitalized developmental psychology, joined by a renewed interest in the psychoses as well as a sociopolitical climate of heightened concern with the lasting negative effects of disadvantage and neglect (Garmezy, 1974: 16). Finally, one should consider the appeal of the promise of rational prevention through targeted interventions (Mosher and Feinsilver, 1971: 19). Even if little followed in terms of new funding and coordinated action on the ground, the 1970s brought an increased focus on prevention, culminating in the report by the Task Force on Prevention submitted to the President’s Commission on Mental Health (1978), and a series of NIMH-led conferences devoted to the topic (Goldstein, 1982; Klein and Goldston, 1977). The idea that programmes should concentrate on high-risk populations was advocated throughout.

High-risk research as a path to primary prevention

While the immediate objective of high-risk research was to capture an empirically verifiable picture of the antecedents of schizophrenic illness, the long-term goal was to promote and refine primary prevention. Since the late 1950s, the state mental hospital resident population had been steadily decreasing – a development facilitated partly by the new tranquilizing drug treatments – but the number of readmissions continued to rise (Mosher and Feinsilver, 1971: 1). Moreover, endless statistics documented what Garmezy (1971: 107) described as ‘the nation’s festering crisis’, especially evident in the impoverished areas of big cities, of general mental ill health and delinquency. Given that treatment was costly and often ineffective, especially once pathology had become entrenched, the movement to create ‘a preventive psychiatry’ gained momentum during the 1960s. However, to many sceptics, the social psychiatric agenda echoed the grandiose schemes of the mental hygiene movement, whose most fervent evangelists had promised the virtual eradication of mental illness through the application of psychological rules for hygienic living (Eisenberg, 1962). In this context of suspicion of weakly conceived, community-wide prevention efforts based on insufficient knowledge, a hybrid research design combining controlled interventions with longitudinal high-risk studies came to be seen as a possible answer to the demand for a more scientific psychiatric prophylaxis. Characterized by specific actions directed at specific populations for specific purposes, such programmes would be based on the best available research evidence and provide shaper definitional clarity along with a formal evaluative design (Garmezy, 1971; Goldston, 1982; Rolf and Harig, 1974). High-risk intervention research was thus described as a potent format for testing theories regarding the central deficits in schizophrenia, and for developing effective strategies to prevent the expression of the illness, or, at least, to lessen its severity.

Nonetheless, intervention studies never came to form a central part of the consortium’s activities. Several possibilities were actively discussed, but mostly on a hypothetical level. The only controlled intervention study was started by Mednick, Schulsinger, and British psychologist Peter Venables (1923–2017) in Mauritius in 1972 on the initiative of the World Health Organization, which stressed the potential benefits of enhancing educational and health services in a developing country (Mednick et al., 1984a). The initial age of the 200 high-risk subjects was only three years, and their selection from a birth cohort of about 1800 children in two Mauritian communities was entirely based not on genetic status but, somewhat controversially, on the type of psychophysiological data obtained in Copenhagen, only this time using more elaborate tests. While acknowledging the difficulty of conducting work in a very different culture and climate, the aim was to examine the utility of deviant electrodermal response to stressful stimuli as a screening tool in a large unselected population, as well as the possibility of prevention through the random assignment of half of the 200 study subjects to special nursery schools where they would receive better nutrition, more physical exercise, cognitive stimulation, and social competence training. In a newspaper article headed ‘Does physiology reveal mental losers?’, Schulsinger stated that they would ‘try to find out if it is possible, already from the kindergarten stage, to forestall the mental disorders in question. It is a unique study, the first truly preventive experiment that has ever been conducted in this field’ ( Søndags-Aktuelt, 1974).

In order to make the intervention more specific, Mednick and Schulsinger were at one point toying with the idea of giving the children small doses of tranquillizers to bring their autonomic functions within the normal range, but this proposition was quickly abandoned (Curran, 1973: 117; Venables et al., 2006: 317). Another important ethical issue, especially in a poor country, was the use of a control group which did not benefit from the nursery experience. Not surprisingly, many of the control parents were dissatisfied with this set-up (Venables, 2014).

Ultimately, very few of the study subjects were found to suffer from a schizophrenic disorder, as defined by DSM-IV. The researchers argued that this might be due to ineffective reporting from the island’s sole mental hospital, or that individuals managed in the community were never diagnosed (Venables, 2014). But by the 1990s the value of the Danish psychophysiological measures had already been questioned by data from other laboratories. The New York and Israeli high-risk projects, for example, found scant evidence of autonomic hyperactivity in high-risk as compared with control subjects and no consistent between-group differences in electrodermal recovery time (Erlenmeyer-Kimling et al., 1985; Kugelmass et al., 1985). A comparison of testing procedures suggested that the high arousal results might stem from the use of unorthodox methods in Copenhagen. In addition, there was the pressing issue of how to make sense of new data revealing a considerable number of electrodermal ‘non-responders’ among high-risk subjects and diagnosed individuals. ¹⁰

Partly as a result of these complications, focus shifted from schizophrenia to schizotypy, antisocial behaviour, and substance abuse, after psychologist Adrian Raine – today best known for his controversial work on the the biological bases of crime – took over the Mauritius project in 1987. Raine and colleagues later found that schizotypic subjects had higher skin conductance responses at an early age, and that the children placed in nurseries ultimately did better on a range of tests and displayed less adult psychopathology, particularly those who were malnourished to begin with (Raine et al., 2010). Interestingly, the current theory is that these outcomes were due partly to the nursery diet, especially the omega 3 in the fish it included. However, demonstrating what some colleagues saw as a tendency to lose interest in ongoing enquiries once new intriguing leads appeared, Mednick’s attention had long before moved elsewhere. Demoralized by the partial collapse of his original theory and the struggle for results, he largely abandoned psychophysiology to concentrate on the role of perinatal complications, including the effects of maternal viral infections (Higgins, 2015; McNeil, 2015).

The fading of the high-risk paradigm

Following a large conference in San Juan in 1984, a major review of all the high-risk projects was prepared by psychologist Norman Watt, who took over as coordinator in 1976. The consortium essentially terminated its work around 1990, though a few studies survived longer. As early as 1976, a review stated that while early enthusiasm had fortunately clouded vision enough to get the studies started, the formidable problems of the high-risk strategy had now become painfully apparent (Keith et al., 1976: 544). Ultimately, despite the vast amount of accumulated data, very little was proven about the conditions thought to create susceptibility and escalate predisposition into full psychosis.

So why did the high-risk programme fall short of expectations? Firstly, researchers faced many challenges due to the longitudinal design and the multiplicity of variables included in the projects. The Risk Research Consortium provided a fine illustration of the multidisciplinarity that characterized schizophrenia research during the 1960s and 1970s. However, this diversity also created incongruities and coordination difficulties. The heterogeneity in samples, measures, procedures, and diagnostic practices both complicated the replication of results and prevented a collaborative approach that would allow data to be pooled across centres in order to achieve greater statistical power (Erlenmeyer-Kimling, 2014; Erlenmeyer-Kimling and Cornblatt, 1987: 459). In addition, the research was expensive, extremely labour-intensive, and required long-term cooperation from subjects. Attrition caused problems in later study phases, as did the fact that the researchers themselves dropped out due to changing interests, death, or retirement.

Continuity in data collection and analysis was furthermore disrupted by changes in the definition and diagnosis of schizophrenia, in aetiological hypotheses, and in testing technologies. The changing conceptualization of the illness itself had important implications for the interpretation of findings. If several more or less connected disorders with varying prognoses were involved, rather than a single illness, one would expect greater phenomenological and aetiological variability. Between the 1960s and 1980s, the Copenhagen investigators thus went from diagnosing subjects dichotomously as schizophrenic or not schizophrenic to distinguishing between different spectrum categories and, using path analysis, between specific ‘longitudinal syndromes’ predicted by different combinations of genetic and environmental influences (Cannon and Mednick, 1993). Moreover, when DSM-III (1980) radically narrowed the diagnostic criteria for schizophrenia, all the original parental diagnoses had to be re-examined (Jørgensen et al., 1987). In the US-based studies that originally employed the broader DSM-II definitions, reclassifications of index parents to bring the diagnoses into line with DSM-III was particularly costly, resulting in a substantial shrinking of the schizophrenia samples (Watt, 2014).

Changing funding patterns also had a significant impact, as primary prevention and longitudinal studies were marginalized in favour of neuroscience, molecular genetics, and the development of biological treatments. In the 1980s, the US federal government restricted funding for research generally, but mainly in the psychosocial realm. By the end of the decade, Watt noted, ‘anyone at the NIMH involved in continuing to promote the consortium, they would be swimming upstream, because we were faced with the genome. It doesn’t have the same level of ambiguity that connected all of us in the risk research consortium. We made a home for biotropes and for sociotropes, and we made them friends to one another’ (Watt, 2014). A related factor was the growing influence of utilitarian attitudes in medical research and the resultant pressure to produce immediate results, which has impelled investigators to adjust their methodology to fit much shorter studies.

In terms of outcomes, a fundamental unresolved issue was the question of specificity. Apparently, most of the reported deviance was not characteristic of schizophrenia alone – similar patterns were found among the offspring of depressed parents on many of the same measures – but seemed more appropriately attributable to unknown factors associated with having a psychiatrically ill parent (Richters and Weintraub, 1990). The cross-diagnostic overlap was also borne out by the many high-risk subjects who developed mental health problems other than schizophrenia. Consequently, the inclusion of multiple psychiatric comparison groups came to be viewed as essential (Neale, 1982).

Attesting to the fundamental uncertainties embedded in the endeavour to characterize and differentiate high-risk children, one major result of the research was, ironically, the rise of the concept of ‘invulnerability’. Researchers soon noticed that many of the children managed well, and that a minority revealed remarkable spirit, high creativity, and leadership, despite having the odds stacked against them. In Copenhagen, psychologist Hanne Schulsinger thus reflected that ‘she often found a certain heroism, a spiritual energy and force in many members of the high risk group’ whereas ‘the low risk group appeared […] more conformist and ordinary’ (Schulsinger, H., 1976: 384). Even in the presence of toxic family and social milieus, outcomes were extremely variable. The most active advocate of the study of resilience was Norman Garmezy (1971: 16). With considerable poetic flair, he argued that the benefits to society of studying the forces that move such children to survival and to adaptation might be far greater than the many efforts at primary prevention centred on notions of vulnerability: ‘With our nation torn by strife between races and between social classes, these “invulnerable” children remain the “keepers of the dream”’. From the mid-1970s, several projects made the concept of competence and other protective factors a key focus, and some risk researchers later joined bodies like the Early Head Start Research Consortium, which aimed to promote general resilience in socio-economically disadvantaged children (Watt, 2014).

Consisting mostly of observational, naturalistic studies led by research psychologists, the risk research programme did not produce much in the way of concrete schemes to actively help exposed youngsters. At a WHO-sponsored conference on the primary prevention of schizophrenia in high-risk groups, held in Copenhagen in 1975, Strömgren (1977) expressed optimism that researchers would soon be able to advise health authorities on effective preventive measures. He imagined that population screenings, perhaps with psychophysiological tests, could be introduced at an early age within the framework of routine child health examinations. However, most speakers urged caution, lest flawed aetiological theories and false expectations should jeopardize future efforts. Only few believed that initiatives like improved obstetric care and the provision of good nursery schools would have a great impact on the incidence of schizophrenia; the hope was still that specific biological causal mechanisms of large effect would soon be discovered, although such developments appeared increasingly remote. Genetic counselling was occasionally mentioned as the best available tool, but the low to moderate risk for offspring rendered it of limited use. Thus, when the NIMH organized a similar conference on prevention five years later, most participants encouraged the use of strategies focused less on deficits and causality than on building strengths and coping skills (Cowen, 1982).

In the end, a major perceived weakness of the genetic risk strategy was that it left the majority of potential schizophrenia cases undetected. Though devised to provide leverage in long-term studies, the eventual yield was relatively sparse, and the generalizability of the findings to the overall population with unaffected close relatives could not be assumed (Gottesman and Erlenmeyer-Kimling, 2001: 99). In view of the enormous problems of accurate prediction, some researchers proposed that early detection of already symptomatic cases by specialized clinical teams might provide the best means of prevention, if one could be reasonably confident that the illness would result from the symptoms (World Health Organization, 1977: 51). Three decades later, this very question proved to be a sticking point in the disputes surrounding the early intervention in psychosis movement represented by figures like Patrick McGorry in Australia and psychoanalyst-turned-biological psychiatrist Thomas McGlashan in the USA. Brimming with therapeutic optimism, it emanated from a clinical psychiatric tradition with little knowledge of, or interest in, the original longitudinal risk studies. The lofty ambitions of uncovering aetiology and facilitating primary prevention were here largely deferred in favour of a more practical activist emphasis on the overriding need for service reform, including better methods of detecting and engaging young people on the very brink of psychosis (Birchwood, 2015; McGorry, 2014). Nonetheless, with conversion rates of only 20 to 30% even in highly selected clinical samples, the problem of false positives persists. ¹¹

Concluding remarks

Tracing the trajectory of the movement for the study of children deemed at high risk for developing schizophrenia, the present article illustrates that current ways of conceptualizing psychosis in terms of risk, and as the culmination of a long developmental process starting sometimes in infancy or even in utero, date back more than half a century. The pioneering Copenhagen High-Risk Project was conceived partly to test a particular theory of schizophrenia as a ‘learned evasion of life’ (Trotter, 1977) in autonomically hyperresponsive individuals, but with time came to emphasize the unexpected discovery of elevated rates of pre- and perinatal insults in sick high-risk subjects. Although a string of negative reports questioned its significance, this finding eventually played a key role in the formulation of the neurodevelopmental hypothesis, which gained increasing acceptance in the field during the 1990s. In fact, by the end of the decade, many researchers in the USA especially went so far as to claim that schizophrenia was simply a neurodevelopmental disorder. Meanwhile, observers associated the stretching of the hypothesis with the new biological zeitgeist. Citing the high frequency of non-psychosis in offspring of two psychotic parents, and the fact that not all schizophrenia cases display brain abnormalities, twin researcher Einar Kringlen criticized leading psychiatrists for ignoring evidence which underlined the importance of non-genetic influences, just like the psychoanalytical establishment had denied the role of biology 40 years earlier (Mjellem and Kringlen, 2001; Kringlen, 2014). Referring to the most deterministic interpretations of the neurodevelopmental model, others spoke of a new Kraepelinian pessimism which, they argued, actually worked against the preventive agenda by distracting interest from the circumstances that trigger the onset of mental illness, and from the planning of pre-emptive interventions (McGorry, 2014).

When contemplating the legacy of the high-risk studies, it is easy to conclude that they failed to deliver the envisioned leap in knowledge that would lead to more effective prevention. While (genetic) high risk emerged as a sturdy research category at the periphery of major psychopathology, its meaning in terms of subjects’ characteristics, experiences, and long-term outcomes remained rather unclear. Risk researchers showed that children of parents with schizophrenia were generally more liable to experience mental disorder, but it was much harder to pinpoint specific, replicable vulnerability markers. Indeed, inspired by the big group of children who managed well against all odds, many investigators instead turned their attention to the role of strengths and protective factors.

From the present-day perspective of disappointment at the unsuccessful search for solid neurobiological markers of schizophrenia, the high-risk pioneers’ dream of reducing incidence through accurate early detection seems almost naïve. Indeed, given current views of schizophrenia as a deeply complex, heterogeneous condition involving at least 100 susceptibility genes – each of tiny effect – it is not surprising that it proved problematical to identify pre-schizophrenics with a single, overall indicator like deviant electrodermal response, which was already difficult to define and measure precisely. Norman Watt (2014) suggests that the biggest achievement of prospective risk research was to humanize mental illness by tangibly highlighting the multitude of influences – including family relationships and functioning – that shape the evolution of psychological abnormality, as well as resilience. Yet, even if gene–environment interactions have come to be seen as central to causality, it is notable that far fewer resources have been spent on the investigation of prevention and environmental risk factors than on genetics and other biological approaches since the 1980s, when the risk consortium terminated its activities.

At any rate, by attempting to determine particular pathogenic mechanisms in samples uncontaminated by the effects of chronic mental illness, Mednick and colleagues made an influential conceptual contribution which still informs the field today (Cornblatt, 2014). Risk researchers also contributed to the shift in emphasis within prevention towards more focused, testable interventions aimed at groups defined as particularly vulnerable to mental illness. In this way, at least, one could argue that they helped pave the way for the current early psychosis framework. But like its predecessor, ‘ultra-high-risk’ research on clinical samples has yet to uncover indicators that can forecast transition to psychosis, and even less so a schizophrenia spectrum disorder, with reasonable precision (Van Os and Guloksuz, 2017). As such, it demonstrates the gulf that continues to separate the hopes and realities of prediction and prevention in psychiatry.