Risk factors for transition to first episode psychosis among individuals with ‘at-risk mental states’

Abstract

Recently developed criteria have been successful at identifying individuals at imminent risk of developing a psychotic disorder, but these criteria lead to 50–60% false positives. This study investigated whether measures of family history, peri-natal complications, premorbid social functioning, premorbid personality, recent life events and current symptoms would be able to improve predictions of psychosis in a group of young, help-seeking individuals who had been identified as being at risk. Individuals (N=74) were followed up at least 1 year after initial assessment. Half the sample went on to develop a psychotic disorder. The most reliable scale-based predictor was the degree of presence of schizotypal personality characteristics. However, individual items assessing odd beliefs/magical thinking, marked impairment in role functioning, blunted or inappropriate affect, anhedonia/asociality and auditory hallucinations were also highly predictive of transition, yielding good sensitivity (84%) and specificity (86%). These predictors are consistent with a picture of poor premorbid functioning that further declines in the period up to transition.

Introduction

Research has revealed a variety of risk factors for the eventual onset of diagnosable psychotic disorders. These include family history, peri-natal complications, premorbid social functioning, premorbid personality, recent life events and current symptoms (Carr et al., 2000). However, the effect sizes of these risk factors is small and their specificity is low so they cannot be used alone in the clinical setting to predict the onset of psychosis. Thus, they have little value for attempts at ‘indicated prevention’ (Mrazek and Haggerty, 1994), that is, the development of interventions designed to delay or even prevent the onset of a diagnosable disorder among those manifesting early, sub-threshold signs of the disorder.

Striking progress towards indicated prevention was made, however, when Yung et al. (1998) developed three sets of operationalised criteria for the identification of help-seeking individuals who are at imminent risk (within 6–12 months) of developing a psychotic disorder, individuals said to be suffering ‘at-risk mental states’ (ARMS). One of these sets comprises sustained but attenuated (subthreshold) psychotic symptoms and another comprises frank (suprathreshold) psychotic symptoms but of limited duration (resolved within 1 week). ‘Trait plus state’ risk factors consist of a first degree relative with a history of any psychotic disorder, or of schizotypal personality disorder, together with any change in mental state or functioning which results in a sustained and significant deterioration in global functioning. In an initial study using these criteria, Yung et al. (1998) found that 8/20 (40%) developed a diagnosable psychotic disorder within 6 months of screening. After increasing this sample and extending the follow-up period, they have reported both that 20/49 (41%) (Yung et al., 2003), and more recently (Yung et al., 2004), 36/104 (35%) developed a psychotic disorder within 12 months. Using similar criteria, but a different method of assessment, Miller et al. (2002b) have reported rates of transition to psychotic disorders, among a group of 13 patients judged to be prodromal, of 46% by 6 months and 54% by 1 year, and a Swedish group has reported that 8 out of 16 became psychotic by 1 year (McGlashan et al., 2001).

This demonstrated ability to identify individuals at imminent risk has persuaded some researchers that preventative interventions in psychotic disorders are a realistic proposition (e.g., McGlashan et al., 2003) and McGorry et al. (2002) have provided evidence that a specific preventative intervention, comprising both low-dose atypical anti-psychotic medication and Cognitive Therapy, maintained over 6 months, is effective at least in delaying the onset of psychosis. Other research groups have also reported trials of this kind that are still in progress Cornblatt et al., 2002, McGlashan et al., 2003. However, ethical objections raised to this kind of preventative strategy DeGrazia, 2001, Schaffner and McGorry, 2001, McGlashan et al., 2001, Bentall and Morrison, 2002 have largely centered on the relatively high false positive rate and consequent unnecessary exposure to treatment.

Such objections to preventative interventions would be lessened if the rate of false positives could be reduced to an acceptable level by identifying additional risk factors for imminent psychosis. A promising step in this direction has been taken by Yung et al. (2003) who followed up 49 individuals meeting ARMS criteria over a 12-month period. They found several significant risk factors for transition, including duration of prodromal symptoms, poor functioning at intake, low-grade positive psychotic symptoms, depression and disorganization, which together provided impressive predictive values between 80% and 90%. However, these were investigated on an ad hoc basis, with cut-offs defined by the data to maximize predictive value. Therefore, these findings require independent cross-validation. Yung et al. (2004) expanded on this sample by adding another 55 participants who met ARMS criteria but, instead of using the additional cases as an independent sample to validate their previously identified risk factors, they combined the samples and selected variables to include in their assessment of predictive ability after examining univariate comparisons between groups (psychotic vs. non-psychotic at follow-up). The baseline variables that they found to be significantly associated with risk of psychosis were membership of both the Trait and Attenuated ARMS groups, long duration of symptoms before referral, severity of depression, reduced concentration and lower global functioning. These variables, once they had been dichotomized, individually all had low sensitivity (<0.35), but a rule requiring the presence of at least one of four predictors led to moderate sensitivity (60%) and high specificity (92.6%).

One of the aims of the present study was to test whether the results published by Yung et al. (2003) could be replicated with an independent sample using data collected by the Psychological Assistance Service (PAS: Carr et al., 2000) of Hunter Mental Health in New South Wales. This is a specialist clinical service, established in 1997, with well-developed referral networks, whose primary function is to identify and treat young people at risk of psychosis.

The assessments conducted by PAS do not include the duration of symptoms. Therefore, it is not possible to investigate all of the predictors identified by Yung et al. However, the assessments do include family history, peri-natal complications, premorbid social functioning, premorbid personality, recent life events and current symptoms. Therefore, it was also possible to test the predictive value of variables within these categories against diagnoses obtained at follow-up.

Since there is evidence of an increased incidence of obstetric complications in those who develop schizophrenia (Geddes et al., 2001), a measure of such complications was included in the PAS assessment protocol. This risk factor has low specificity in the general population, but its specificity might be much higher among help-seeking individuals who meet ARMS criteria. Poor premorbid social functioning has long been associated with schizophrenia but now its status as a risk factor has been confirmed in population-based cohort studies Jones et al., 1994, Done et al., 1994, Malmberg et al., 1998, Davidson et al., 1999. Differences in social adjustment between controls and children who went on to develop psychotic disorders have been found as early as 7 years of age Done et al., 1994, Jones et al., 1994, though the differences tend to be larger at age 11 (Done et al., 1994) and 16–17 (Davidson et al., 1999). Therefore, separate measures of premorbid social adjustment in childhood and early adolescence were included in the present study. Since adolescents who eventually develop psychotic disorders have also been found to differ from matched controls in intelligence David et al., 1997, Davidson et al., 1999, a measure of premorbid intelligence was also included.

It has also been found in prospective studies that schizotypal characteristics are a risk factor for psychotic disorders. Using data from the Edinburgh High-Risk Study, Miller et al. (2002a) examined the ability of schizotypal characteristics to predict onset of schizophrenia among young people who had at least two first- or second-degree relatives with schizophrenia. The participants were assessed with a structured interview for schizotypy, their responses were subjected to a principal components analysis, and the four components that emerged from this analysis were used to predict onset of schizophrenia. It was found that 7 of a sample of 78 high-risk individuals who were followed up for 39 months developed schizophrenia and these could be discriminated prospectively from their scores on the schizotypal components with a positive predictive power of 0.67 and a negative predictive power of 0.99. Social withdrawal was the best single predictor but psychotic symptoms, socio-emotional dysfunction and odd behavior also made independent contributions. Therefore, a measure of schizotypal characteristics was also included in the present study.

The criteria for ‘At-Risk Mental States’ that Yung et al. (1998) developed are largely based on current frank and attenuated positive symptoms of psychosis. However, there are reasons to believe current attenuated negative symptoms are also an important part of the prodrome of psychosis, as Cornblatt et al. (2002) have argued. For example, Kwapil (1998) identified 34 college undergraduates who scored particularly highly on a questionnaire measure of social anhedonia and found that, at 10-year follow-up, 24% of them had developed a schizophrenia-spectrum disorder (compared with only 1% of a control group). There is evidence for a significant association between substance abuse (Buhler et al., 2002), especially cannabis abuse (Degenhardt, 2003), and later development of psychotic symptoms. There is also evidence, albeit from a retrospective study, that an excess of stressful life events precedes the onset of psychoses (Bebbington et al., 1993). Therefore, measures of current negative symptoms, in addition to current positive symptoms, substance abuse and stressful life events were also included in the present study.

In summary, the aim of the present study was to test whether it is possible to improve on the ability of the ARMS criteria to predict which help-seeking individuals will develop a psychotic disorder. Three separate analyses were planned. One of these tested whether the additional risk factors identified by Yung et al. (2003) could be replicated. Another analysis tested whether other variables which have been identified as risk factors for psychosis would enhance the ARMS criteria: family history, peri-natal complications, premorbid social functioning, premorbid personality, recent life events and current symptomatology. Many of these risk factors refer to times quite remote from the present. Therefore, although they may have influenced individuals to develop the kind of problems that are included in the ARMS criteria, they may have little power to discriminate, among those who meet these criteria, between those who are truly in a prodromal phase and those who are not. It seems likely that current problems will have more power for such discrimination. Thus, the third analysis examined whether current symptoms and functioning, and recent life events, could identify which individuals were truly in a psychotic prodrome.