Identifying a system of predominant negative symptoms: Network analysis of three randomized clinical trials

Abstract

Background

Reasons for the recent mixed success of research into negative symptoms may be informed by conceptualizing negative symptoms as a system that is identifiable from network analysis. We aimed to identify: (I) negative symptom systems; (I) central negative symptoms within each system; and (III) differences between the systems, based on network analysis of negative symptoms for baseline, endpoint and change.

Methods

Patients with chronic schizophrenia and predominant negative symptoms participated in three clinical trials that compared placebo and amisulpride to 60 days (n = 487). Networks analyses were computed from the Scale for the Assessment of Negative Symptoms (SANS) scores for baseline and endpoint for severity, and estimated change based on mixed models. Central symptoms to each network were identified. The networks were contrasted for connectivity with permutation tests.

Results

Network analysis showed that the baseline and endpoint symptom severity systems formed symptom groups of Affect, Poor responsiveness, Lack of interest, and Apathy-inattentiveness. The baseline and endpoint networks did not significantly differ in terms of connectivity, but both significantly (P < 0.05) differed to the change network. In the change network the apathy-inattentiveness symptom group split into three other groups. The most central symptoms were Decreased Spontaneous Movements at baseline and endpoint, and Poverty of Speech for estimated change.

Conclusions

Results provide preliminary evidence for: (I) a replicable negative symptom severity system; and (II) symptoms with high centrality (e.g., Decreased Spontaneous Movement), that may be future treatment targets following replication to ensure the curent results generalize to other samples.

Introduction

Since Kraepelin's historic portrayal of the destruction of the personality (Kraepelin, 1971), negative symptoms have been considered as central to schizophrenia. Negative symptoms are associated with deficits in cognitive, social and real-world functioning (Bowie et al., 2006, Harvey et al., 2006, Kirkpatrick et al., 2006), and are more relevant to functioning than positive symptoms (Rabinowitz et al., 2012). Despite being relevant to functioning in schizophrenia, meta-analysis has established that second-generation antipsychotic medications have efficacy in the treatment of positive and not negative symptoms of schizophrenia disorder (Leucht et al., 2009). Furthermore, at present no treatment for negative symptoms has attained the clinically significant improvement threshold (Fusar-Poli et al., 2015). The inefficacy of medication to treat negative symptoms led to a NIMH-MATRICS expert consensus group statement on negative symptoms (Alphs, 2006, Kirkpatrick et al., 2006, Kirkpatrick and Fischer, 2006, Marder et al., 2011). That group has, for instance, highlighted methodological and assessment limitations in clinical trials of negative symptoms (Kirkpatrick et al., 2006). Despite consensus surrounding negative symptoms, subsequent treatment initiatives for negative symptoms have failed to demonstrate efficacy (e.g., biopertine, mGlu2/3). One reason for the mixed success of these initiatives may be the conceptualization, measurement and derivation of treatment targets for negative symptoms.

Generally to conceptualize and understand the nature of negative symptoms studies have used factor analysis. For instance, factor analytic studies have examined Scale for the Assessment of Negative Symptoms (SANS; Andreasen, 1983) since it is the most widely used and approved negative symptom measure (Kirkpatrick et al., 2006). In theory, the SANS is comprised of five a priori symptom factors (i.e., symptom clusters) of affective flattening, alogia, avolition, anhedonia. However, factor analytic studies that examine the nature of negative symptoms provide inconsistent results. Studies have identified that the SANS comprises of two (Toomey et al., 1997), three (Keefe et al., 1992, Kelley et al., 1999, Levine and Leucht, 2013b, Mueser et al., 1994, Sayers et al., 1996), four (Rabany et al., 2011), and five (Peralta et al., 1995) different symptom factor structures (i.e., co-varying symptom clusters). Hence there appears to be a lack of consistent identification of the nature of negative symptoms.

Existing understanding and hence identification of treatment targets for negative symptoms may be elaborated on beyond factor analysis by using network analysis for several reasons. We summarize the key differences between network and factor analyses in Table 1. First, in factor analysis an unmeasured (i.e., implicit or technically ‘latent’) symptom factor ‘causes’ the associations among the observed symptom rating scores. Seen this way, for example, a proclivity to inattentiveness implicit symptom factor ‘causes’ the SANS symptoms of test and social inattentiveness. It is noted that the proclivity to inattentiveness is not measured per se, rather it is an explicit latent variable in factor analysis. Unlike factor analysis, network analysis considers negative symptoms as a system. Seen this way, for example, test and social disturbances group together. The association between two symptoms cannot be attributed to another symptom, but they may be legitimately associated with some of the remaining negative symptoms. Second, factor analysis cannot ascertain which of the negative symptoms are central. This is in contrast to network analysis where the extent to which a negative symptom connects with the remaining symptoms in the network varies both by the number and magnitude of connections with the other negative symptoms in the system. This conceptualization of negative symptoms as a network is reminiscent to current notions regarding connectivity between brain circuits in neuropsychiatry.

To understand the negative symptom system we apply network analysis to baseline, endpoint and change SANS items in three clinical trials of predominant negative symptoms. We aim to identify negative symptom networks, the most central negative symptoms within each symptom network, and differences between networks.