Health outcomes in people at high risk for psychosis who do not develop the illness: a meta-analysis

Indicated prevention in individuals at clinical high risk for psychosis (CHR-P) is one of the most promising primary preventive approaches in psychiatry (Fusar-Poli et al., 2017b). CHR-P individuals are young and they accumulate risk factors such as living alone or being unemployed (Fusar-Poli et al., 2017e; Radua et al., 2018; Oliver et al., 2019) that enrich their level of risk for psychosis (Fusar-Poli et al., 2016c). In turn, this can lead to functional impairments (Fusar-Poli et al., 2015c) and the onset of attenuated psychotic symptoms (Fusar-Poli et al., 2017c). The distress associated with these experiences can prompt CHR-P individuals to seek help (Falkenberg et al., 2015) at specialised mental health clinics (Kotlicka-Antczak et al., 2020; Salazar de Pablo et al., 2021a). In these clinics, the prognosis is formulated reaching very good accuracy using psychometric instruments (Fusar-Poli et al., 2015a).

The majority of CHR-P individuals do not transition to psychosis within the first 2 years of presentation. After 2 years, 16% of CHR-P individuals transition to psychosis, and the transition risk continues to rise until about 4 years of follow-up, reaching 36% at 10–11 years (Salazar de Pablo et al., 2021c). Clinical outcomes in CHR-P individuals who do not transition to psychosis remain scattered, heterogeneous and inconsistent (Simon et al., 2011). It remains unclear how many will improve, permanently or only temporarily (Schultze-Lutter, 2009; Fusar-Poli et al., 2013), or whether the poor mental health outcomes observed are mostly driven by the presence of transitioning CHR-P individuals. Longitudinal research comparing individuals who develop psychosis with those who do not is overall inconsistent (Fusar-Poli et al., 2013).

Following an earlier meta-analysis, published 10 years ago, evaluating the proportion of individuals who do not transition to psychosis and the frequency of remission (Simon et al., 2011), a more recent update found that 28–71% of CHR-P individuals who do not transition to psychosis, do not remit from their CHR-P either and 22–82% still have one or more clinical diagnosis in the long-term (Beck et al., 2019a). Additional meta-analyses have focused on other clinical outcomes in all CHR-P individuals (transitioning and non-transitioning individuals), including the level of functioning (Fusar-Poli et al., 2015c, 2017e), quality of life (Fusar-Poli et al., 2015c), comorbid disorders (Albert et al., 2018) and remission (Simon et al., 2013). However, no comprehensive meta-analysis has addressed these outcomes together in individuals at CHR-P not developing psychosis. Also, no meta-analysis has evaluated outcomes in CHR-P individuals not transitioning to psychosis, comparing to those developing it. We aimed to comprehensively assess the broad longitudinal clinical outcomes of attenuated psychotic symptoms, negative symptoms, depressive symptoms, functioning and remission in CHR-P individuals who did not transition to psychosis. Furthermore, we aimed to compare these outcomes between those CHR-P transitioning or not to psychosis, while controlling for some potential moderators.

The protocol for this study was registered on PROSPERO (CRD42021229212). This study was conducted in accordance with the ‘Preferred Reporting Items for Systematic reviews and Meta-Analyses’ (PRISMA) (Moher et al., 2009) (online Supplementary eTable 1), ‘Meta-analysis Of Observational Studies in Epidemiology’ (MOOSE) (Stroup et al., 2000) (online Supplementary eTable 2) and ‘Reporting Tool for Practice Guidelines in Health Care’ (RIGHT) (Chen et al., 2017) statements.

To our knowledge, this is the first meta-analysis to comprehensively evaluate numerous clinical outcomes (attenuated psychotic symptoms, negative symptoms, depressive symptoms, functioning and remission) in CHR-P individuals who do not transition to psychosis. Evaluating 2756 CHR-P individuals from 23 prospective cohorts, we showed that, although CHR-P individuals improved on several outcomes, more than half of them did not reach remission.

The main finding of this meta-analysis is to have demonstrated a variable improvement of clinical outcomes over follow-up. The effect size for the improvement in the evaluated outcomes was large for attenuated psychotic symptoms (Hedges’ g = 1.410) and depressive symptoms (Hedges’ g = 0.844), and medium for negative psychotic symptoms (Hedges’ g = 0.683). Furthermore, there was a medium effect size for functional improvements (Hedges’ g = 0.776). The large effect size improvements in attenuated psychotic symptoms compared to other outcomes may be explained by a better therapeutic response for positive symptoms, which tends to respond better to antipsychotic medication or cognitive-behavioural therapy, compared to other outcomes such as negative symptoms (Woodward et al., 2014). However, there is no convincing evidence suggesting that specific preventive interventions can specifically improve attenuated (as opposed to established) psychotic symptoms better than others, including needs-based interventions (Davies et al., 2018; Fusar-Poli et al., 2019a, 2020c). A specific concern relates to the impact of antipsychotics in CHR-P individuals, which appears to be largely secondary to the high comorbidities accumulated and their transdiagnostic impact on the clinical presentation (Salazar de Pablo et al., 2020b; Fusar-Poli and Salazar de Pablo, 2021). Similarly, depressive symptoms seem to improve over time, which is relevant clinically given that depression and anxiety are common reasons for seeking help in CHR-P individuals (Falkenberg et al., 2015). Furthermore, previous meta-analytical evidence suggests that 40.7% of CHR-P individuals (Fusar-Poli et al., 2014) and 49% of individuals with DSM-5 Attenuated Psychosis Syndrome (Salazar de Pablo et al., 2019) present with comorbid depressive disorders. Previous studies have indicated that persistence of depression in CHR-P individuals is associated with decreased remission from a CHR-P status (Rutigliano et al., 2016; Kline et al., 2018; Fusar-Poli et al., 2019c). Depression is also associated with more pronounced negative psychotic symptoms and general symptoms, and it may contribute beyond the impact of positive and negative symptoms to impairments in social functioning (Fusar-Poli et al., 2014). Despite these findings, depressive symptoms do not lead to an increased risk of developing psychosis in CHR-P individuals (Fusar-Poli et al., 2014). Improvements on negative findings were less marked but still significant. We also showed that negative symptom improvements in CHR-P individuals who do not transition are of similar magnitude to those observed in individuals with schizophrenia (ES = 0.66) (Savill et al., 2015). Negative symptoms are often the first symptoms that CHR-P individuals develop (Metzak et al., 2020). There is a strong negative relationship between negative symptoms and functioning in CHR-P individuals (Devoe et al., 2020; Metzak et al., 2020), a finding also confirmed in individuals with schizophrenia (Ventura et al., 2009). Therefore, the observed clinical improvements in the severity of core CHR-P symptoms were paralleled by functional improvements at follow-up. It is well established that, similar to other psychiatric disorders, functional impairments are common in CHR-P individuals (Fusar-Poli et al., 2015c). However, the functional improvement may not be sufficient to reach the full functional remission (see below). Previous evidence on CHR-P individuals not transitioning to psychosis indicated that 45.3% of them still remain functionally impaired after 6 years (Rutigliano et al., 2016). This is not surprising given the lack of robust interventions to improve functional outcomes in this population.

Another core finding of this meta-analysis is to have complemented the analysis of continuous outcomes such as the severity of symptoms or levels of functioning with other real-world categorical outcomes that are directly informative of clinical care. In fact, statistically significant improvements of attenuated psychotic symptoms, depressive or negative symptoms and functioning do not automatically translate into tangible benefits for the lives of CHR-P individuals. This phenomenon has already been observed in psychopharmacological interventions for negative symptoms in schizophrenia, where statistically significant improvements were associated with negligible patient-level perceived improvements (Fusar-Poli et al., 2015b). Indeed, despite the symptomatic and functional improvements observed above, our meta-analytic frequency of remission indicates that only less than half (48.7%) of CHR-P not developing psychosis eventually remitted at follow-up. Our results align with a systematic review which found that 28–71% of CHR-P individuals who do not transition to psychosis, do not achieve remission (Beck et al., 2019a). Our increased frequencies of remission are due to the exclusion of CHR-P individuals who transitioned to psychosis. Interestingly, our frequency of remitters appeared lower than that observed during a first episode of psychosis (58% at 66 months) (Lally et al., 2017). This conflicts with one of the core foundations of the clinical staging model, which assumes that early stages are associated with a more favourable outcome and the likelihood of remission (Fusar-Poli et al., 2017b). Future research is required to address this issue.

According to our sensitivity analyses, frequencies of remission do not substantially increase throughout the follow-up. Improvements in functioning, attenuated psychotic symptoms and negative symptoms are also variable, without a significant pattern towards a prolonged improvement. This suggests that preventive interventions and monitoring may be needed in the long-term to support CHR-P individuals who do not remit, and individuals with negative symptoms and poor functioning. Currently, only 27.6% of clinical services to prevent psychosis provide care for more than 24 months (Salazar de Pablo et al., 2021a). Unfortunately, 2 years of care is not sufficient to capture the very real long-term clinical outcomes of this vulnerable population (Fusar-Poli et al., 2020b). A need for specialised services to detect CHR-P individuals and to offer needs-based and psychological interventions has been identified (Fusar-Poli et al., 2020d). This meta-analysis advances knowledge by clarifying that those CHR-P individuals who do not transition require extended support for their mental health, as non-transition does not automatically imply restoring a healthy status.

Our secondary analyses compared CHR-P individuals developing psychosis and not on the same outcomes. We found that functioning (Hedges’ g = 0.623) and attenuated psychotic symptoms (Hedges’ g = 0.706) improved in CHR-P individuals who did not transition to psychosis compared to those who did. Functioning is closely related to both the duration and severity of attenuated psychotic symptoms (Salazar de Pablo et al., 2020a). These findings indicate that the level of functioning of CHR-P individuals is strictly closed to transition to psychosis (Fusar-Poli et al., 2015c), confirming that transition to psychosis from a CHR-P state is associated with severe real-world clinical outcomes. A recent study confirmed that CHR-P individuals transitioning to psychosis (n = 130) were more likely to receive antipsychotic medication, to be admitted informally and on a compulsory basis, and to have spent more time in hospital than first-episode patients who presented to early intervention services (n = 1121), with a comparable likelihood of receiving clozapine (Fusar-Poli et al., 2020a). Surprisingly, we found no statistically significant differences in remission, negative and depressive symptoms between CHR-P individuals who transition or not to psychosis. It is important to highlight that statistical potency for these analyses was low, due to the limited number of prospective cohorts included. The lack of differences on negative or depressive features may be explained by the fact that no effective interventions are available for these domains in CHR-P individuals, beyond the presence of subthreshold or frank positive psychotic symptoms. The lack of difference in frequency of remission may be explained by the rapid and intensive package of early intervention care typically received by CHR-P individuals who develop over threshold psychosis, which has been demonstrated to be highly effective in improving clinical outcomes (Correll et al., 2018).

According to the meta-regressor factors tested, higher age was associated with a higher longitudinal improvement in attenuated psychotic symptoms. We previously found that age does not appear to modulate transition risk (Catalan et al., 2020). However, positive psychotic symptoms, such as hallucinations, are common in children and adolescents (Stevens et al., 2014). This may be related to the high prevalence of previous traumatic events in adolescents at CHR-P (63.2–85.0%) (Catalan et al., 2020), which are associated with the severity of positive symptoms, such as hallucinations and delusions (Bailey et al., 2018). There is also converging evidence indicating that early onset of psychosis is associated with poorer longer-term outcomes (Salazar de Pablo et al., 2021b). We also found that publication year was associated with better functioning. While declining transition risks over the years have been observed in the early days (Fusar-Poli et al., 2012), these have been disconfirmed in the most recent literature (Salazar de Pablo et al., 2021c). These discrepancies may be due to heterogeneous risk enrichment during the recruitment of young adults undergoing a CHR-P assessment (Fusar-Poli et al., 2016c, 2016d; Rice et al., 2019), with the associated variable impact of comorbid conditions (Salazar de Pablo et al., 2020a). Overall, the association between publication year and functioning may be secondary to sampling biases (Fusar-Poli et al., 2016d, 2018). We also found lower frequencies of remission in BLIPS individuals. This is an established finding, because BLIPS have poor mental health outcomes (Fusar-Poli et al., 2017b, 2017d) such as the higher risk of developing psychosis, being admitted compulsorily into hospital, receiving antipsychotics and benzodiazepines and lower probability of receiving psychotherapy (Fusar-Poli et al., 2016a, 2016b, 2017a, 2017d, 2019b, 2020b).

This study has several limitations. First, the available studies to meta-analyse some of the clinical outcomes (e.g., depressive and negative symptoms) evaluated was limited. As noted above, lack of power for some may have led to non-statistically significant differences in frequency of remission between CHR-P individuals who transitioned to psychosis or not. However, the database was sufficiently powered to run our primary analyses. Second, additional outcomes, such as quality of life, were not assessed as these were hardly ever reported in a meta-analysable manner. Third, there was high heterogeneity across the included studies, which we partially addressed in meta-regression analyses. Fourth, it was not possible to test some meta-regressors, including duration of untreated attenuated psychotic symptoms, ICD or DSM-defined comorbidity and exposure to baseline interventions, due to a limited amount of studies. Fifth, as noted above, the clinical meaning of changes in clinical and functional outcomes are not always directly interpretable. For instance, a decrease from 6 to 3 in one of the CHR-P symptoms may be more clinically relevant than a decrease from 3 to 0. Nevertheless, these findings confirm that transition to psychosis is not a trivial event of little clinical meaning, as argued by some authors, but related to real-world morbidity and mortality (Fusar-Poli et al., 2020b). Sixth, as transition to psychosis is defined by the worsening of attenuated psychotic symptoms, differences in secondary outcome were somewhat tautologically expected. Finally, we could not stratify our results according to the CHR-P subgroups, and we could only test their association with one of our outcomes (remission).

Clinical outcomes improve in CHR-P individuals not transitioning to psychosis but only less than half remit over time. Sustained clinical attention should be provided in the longer term to monitor these outcomes.