Long-term trajectories of positive and negative symptoms in first episode psychosis: A 10 year follow-up study in the OPUS cohort

Abstract

Objective

Knowledge about course of illness can help clinicians to develop effective interventions and improve treatment outcomes. The goal of this study was to construct positive and negative symptom trajectories based on structured clinical assessments collected over 10 years within a cohort of people with first episode psychosis.

Method

A cohort of 496 people with first episode psychosis (ICD-10, F20-28) originally recruited for the OPUS study (1998–2000) and treated in community psychiatric services were rated on clinical symptoms at 5 different occasions across ten years. Psychopathology was assessed using the Scales for Assessment of Positive and Negative Symptoms. Symptom trajectories were constructed using Latent Class Analysis.

Results

Five distinct trajectories were identified for positive symptoms (response — 47%, delayed response — 12%, relapse — 15%, non-response — 13% and episodic response — 13%). Four distinct trajectories were identified for negative symptoms (response — 28%, delayed response — 19%, relapse — 26% and non-response — 27%). Multivariable regression analysis of baseline characteristics identified that longer duration of untreated psychosis (OR 1.27–1.47, p < 0.05) and substance abuse (OR 3.47–5.90, p < 0.01) were associated with poorer positive symptom trajectories (higher levels of psychotic symptoms) while poor social functioning (OR 1.34–5.55, p < 0.05), disorganized symptoms (OR 2.01–2.38, p < 0.05) and schizophrenia diagnosis (OR 5.70–8.86, p < 0.05) were associated with poorer negative symptom trajectories (higher levels of negative symptoms). A proportion of people displayed significant changes in symptoms several years after diagnosis.

Conclusions

Trajectories of illness for positive and negative symptoms were heterogeneous among people with first episode psychosis. Positive symptoms showed a general pattern of reduction and stabilization over time while negative symptoms typically showed less variation over the ten years. Results have implications for the focus, timing and length of interventions in first episode psychosis.

Introduction

The investigation of long-term outcomes is a key focus within schizophrenia research. Information on long term prognosis can assist in the identification of factors that could impact on illness course. While it has been previously proposed that the typical course of illness in schizophrenia is chronic and deteriorating (Davidson and McGlashan, 1997), recent studies have indicated that a heterogeneous course of illness is most characteristic, with a significant number of people achieving symptom remission and recovery (Van Os and Kapur, 2009, Henry et al., 2010a, Harrow et al., 2005, Wunderink et al., 2009).

Traditionally, long-term outcomes in schizophrenia have been assessed by investigating the rates of symptom remission and recovery at a specified time after diagnosis (Menezes et al., 2006). While these studies provide valuable cross sectional information about levels of psychopathology and functioning for people with schizophrenia, they do not provide detailed information about how the illness is manifested over time. Studies using longitudinal data can contribute to a better understanding of the course of illness and help clinicians to adjust the timing and intensity of interventions to optimize treatment outcomes for people with schizophrenia.

The classification of outcomes into dichotomous categories such as recovered or not and symptom remission or not, is commonplace within schizophrenia research but this practice can also be problematic. If dichotomization is undertaken without evidence of a bimodal distribution, this may lead to the inefficient analysis of continuous data, where up to one third of the data from the sample is thrown away (Streiner, 2002, Royston et al., 2006). Furthermore, cut-off scores for membership into a particular category are often arbitrary in nature (Mulder et al., 2003). The oversimplification of complex outcomes into categories such as recovered/not recovered can create difficulties in the translation of results into clinically meaning information (Rietschel et al., 1999).

Advances in statistical modeling techniques have meant that it is now possible to identify categories based on temporal patterns of change using latent class models (LCA) and growth mixture modeling (GMM) (Beunckens et al., 2008). The advantage of these methods is that the subpopulations and thresholds do not need to be defined or identified arbitrarily. Furthermore, LCA and GMM are sensitive to the pattern of change over time and are robust in the presence of missing data (Muthen et al., 2002, Muthen and Brown, 2009).

Recently, a number of studies have used LCA or similar growth mixture models to identify different trajectories to treatment response within schizophrenia. Rabinowitz and colleagues identified three response trajectories within a cohort of people with schizophrenia over a period of ten years. A significant majority showed improvement and stabilization in positive symptoms (Rabinowitz et al., 2007). Levine et al. (2011) identified five distinct illness trajectories, where the typical course was one of initial deterioration followed by progressive amelioration (Levine et al., 2011). A limitation of both these studies was the use of number of days hospitalized as a proxy for clinical symptoms, which may not directly reflect levels of psychopathology, given the criteria for admission to hospital can vary depending on the health service.

A number of studies have identified trajectories based on clinical ratings of positive symptoms. Schennach et al. (2012) identified five response trajectories, where the majority of participants showed a moderate to good response (Schennach et al., 2012) while Case et al. (2010) identified four trajectories with over 80% of participants displaying a pattern of gradual improvement within the first three months of treatment (Case et al., 2010). The results from these studies were limited due to the very short follow-up period (12 weeks), given that schizophrenia is an illness manifested over several years. While these modern statistical techniques are generating new information about symptom trajectories in schizophrenia, further longitudinal studies using clinical data are required. Furthermore, there are no current studies that have examined negative symptom trajectories using clinical data and latent class analysis.

The aim of this study was to identify discrete trajectories for positive and negative symptoms using structured clinical assessments collected over a ten year period within a large representative sample of people with first episode psychosis. Additionally, the study identified which baseline characteristics discriminated between different trajectories.