Introduction
According to the latest estimates of the World Health Organization, schizophrenia affects approximately 1 in 300 people or 24 million people worldwide [
1], with the annual incidence being steadily on the rise in certain regions [
2,
3]. Although it is not as common as many other mental disorders, and despite recent advances in psychiatry, it remains one of the most debilitating disorders. Indeed, schizophrenia is associated with high rates of suicidality [
4], excess early mortality [
5], and impaired quality of life [
6]. In addition, existing evidence indicates that only nearly one in five patients with schizophrenia report achieving clinical recovery [
7], and one in seven will achieve functional recovery [
8]. This variability in outcomes was partly attributable to substantial and hardly to delineate clinical heterogeneity of schizophrenia [
9]. Several researchers proposed that schizophrenia represents a heterogeneous disorder with a group of distinct symptom structures and overlapping clinical phenotypes, rather than a unitary disease entity [
10‐
12].
Clinical heterogeneity of schizophrenia
The clinical heterogeneity of schizophrenia manifests in different symptom patterns, course trajectories, and treatment responses [
13,
14]. Interestingly, these assumptions date back to Kraepelin’ s view of dementia praecox as “the expression of a single morbid process, though outwardly they often diverge very far from one another” [
15], and Bleuler’ s labelling as “Group of Schizophrenias” [
16]. Since the earliest conceptualizations of schizophrenia, there have been multiple efforts to develop subtypes that may help advance our understanding of the disease and its underlying pathogenesis [
17‐
21]. The identification of schizophrenia subgroups also has many benefits in terms of informing “how to treat” based on a set of treatment choices (in contrast with the dimensional approach that rather informs the decision “to treat” or “not to treat”), and predicting response to treatment above and beyond symptoms severity and extent of suffering or disability [
22]. However, traditional attempts to capture and map clinical heterogeneity of schizophrenia were unsuccessful and hampered by the lack of temporally stable and clinically valid taxonomic schemes. The subtyping scheme previously proposed in the Diagnostic and Statistical Manual of Mental Disorders (DSM) was dropped by the fifth edition because of its limited clinical validity, prognostic value, and research utility [
23‐
25]. Researchers have, for example, individualized a putative “deficit” subtype with primary and persistent negative symptoms (NS) characterized by distinctive pathophysiology, clinical course, and treatment response [
18,
26]. Although deficit schizophrenia seems to represent “the most viable candidate for a separable disease entity” within the disease [
27], the categorical approach of NS is not uniformly accepted and some researchers lean more toward a dimensional approach where patients differ in the “amount” or “degree” rather than “status” [
28].
In the DSM-5, NS are defined as one of the core dimensions of schizophrenia [
29]. NS are common, with at least one NS being reported by over 50% of people with schizophrenia [
30,
31]. They reflect an absence or reduction of functions that are normally present in the general population, and that are either related to expressive functions (e.g., alogia, blunted affect) or motivation and interest (e.g., asociality, anhedonia, avolition). This group of symptoms have greater impact on functioning than positive symptoms [
32], and poses a significant burden on patients, their families, and the healthcare system. However, NS remain neglected targets for treatment [
33], with most attention having rather been directed at treating positive symptoms that are more easily detectable and manageable [
34]. Overall, the lack of therapeutic advances and mixed research findings about the underlying mechanisms of the negative symptomatic dimension of schizophrenia have emphasized the need for their reconceptualization [
35]. Over the last years, considerable progress has been made in this direction. Studies have begun to acknowledge and provide valuable insight into the heterogeneity of NS (e.g., differentiating persistent vs. transient and primary vs. secondary NS) [
36‐
38].
Previous cluster analyses of NS
To better characterize the variable phenotypic expression of symptoms, Goldstein [
39,
40] called for a clinically-informed and statistically supported nosology by using multivariate analysis (i.e. Cluster Analysis [
41]) that enables identifying homogenous subgroups of patients within the broader schizophrenia diagnosis. To answer this call, numerous studies have applied this data-driven statistical approach to dissect the heterogeneity of schizophrenia. A recent systematic review of cluster and group-based studies showed that most of the existing studies addressed cognitive deficits, and could only find four cross-sectional studies assessing NS clusters [
42]. These four studies depicted two to five patient trajectory groups, and identified a number of sociodemographic (e.g., gender, age, educational status, marital status) and clinical (e.g., duration of untreated psychosis, social functioning, quality of life, cognitive performance, depressive symptoms) predictors [
42]. Among the four studies, two investigated NS in combination with positive symptoms and found three to four clusters (i.e., low scores on both positive and negative symptoms, high scores on both sets of symptoms, predominantly negative or positive symptoms scores) [
43,
44]. In a different approach and method by Strauss et al. [
45], NS were regarded as two separate factors instead of unitary NS scores and subsequently demonstrated two distinct symptom clusters, one characterized by predominant diminished expression while the other by higher anhedonia and avolition symptoms. The latter subgroup was marked by a poorer clinical and functional status [
45].
For a better generalizability and a more optimal conceptualization of NS in schizophrenia, some studies have recently provided support for a latent five-factor model (i.e., blunted affect, alogia, anhedonia, asociality, and avolition) [
46‐
49] that should be favored over both the one- and two-factor models [
50]. This five-factor model of NS has initially been proposed in a consensus development conference held by the National Institute of Mental Health [
51], and later its construct validity was consistently proven robust across different languages and cultures (e.g [
47‐
49].,). Surprisingly however, the five NIMH consensus NS domains have only very recently been entered into a cluster analysis in a first and only study by Paul et al. [
52]. Authors used a semi-structured interview designed to assess the five consensus domains (i.e., the Brief Negative Symptom Scale [
53]), and could identify four distinct clusters within a sample of people diagnosed with schizophrenia or schizoaffective disorder: (1) low NS, (2), severe NS, (3) moderate NS with predominantly elevated blunted affect, and (4) moderate NS dominated by avolition [
52]. The four clusters differed in their relationships with neurocognition, clinical characteristics, and functional outcome, thus suggesting that the NS subgroups established have distinct clinical and neuropsychological profiles [
52].
Rationale of the present study
In the present study, we sought to build on and extend prior research by providing some new insight into potential and unique NS subgroups within schizophrenia. We contribute to the existing literature in two ways. First, we respond to recent calls for research entering the latest five NIMH consensus NS domains into a cluster analysis in schizophrenia. We further add to the literature by using for the first time a self-report measure, i.e. the Self-Evaluation of Negative Symptoms (SNS) [
54], in contrast with the previously used semi-structured interview (i.e. Brief Negative Symptom Scale (BNSS) [
53]). There is some evidence on discrepancies between examiner-ratings and self-ratings for NS, with a majority of patients found to be unable to accurately report their symptoms in hetero-evaluation [
55,
56]. In this regard, self-evaluation was proven to detect some clinical information on NS recognized and analyzed by the patients themselves that are not necessarily captured by clinicians in a standard interview [
57]. Additionally, the SNS has proven to enable patients to express their speech expression, loss of emotion, and deficits in motivation regardless of their depressive symptoms [
54,
58,
59]. Second, we investigate for the first time in this area a clinical population from an Arab country of the Middle East and North African (MENA) region. A recent systematic review revealed that the vast majority of studies available on clusters of schizophrenia symptoms were conducted in the USA and/or other Western countries [
42]. Even though schizophrenia is a universal disease, affecting people all around the world regardless of origin, race, or culture social class [
60], recent research suggests that it manifests differently across countries. Indeed, several studies pointed to marked variations in the expression and clinical trajectory of schizophrenia symptoms (including anhedonia, depressive symptoms, and emotional processing) depending on sociocultural, ethnic and contextual factors [
61‐
64]. Hence, it is pertinent to provide the first data on the topic from a non-Western developing country of the largely under-researched MENA region, Lebanon. Therefore, our main objectives were: (1) to use cluster analysis to identify subgroups of Lebanese patients diagnosed with either schizophrenia or schizoaffective disorder based on NS clusters, and (2) to relate the statistically-derived subgroups to clinically relevant external validators (including measures if state and trait depression, stigma, insight, loneliness, social support). We hypothesized that clustering would allow us to accurately identify subgroups of patients characterized by severe and low levels of NS, and another intermediate subgroup with moderate levels of NS.
Discussion
Dissecting the heterogeneity of schizophrenia may help foster progress in understanding its etiology and lay the groundwork for the development of new treatment options for primary or enduring NS. In this regard, the present study adopted a cluster analysis approach to classify patients with schizophrenia and schizoaffective disorder based on the five NS domains social withdrawal, emotional withdrawal, alogia, avolition and anhedonia. A three-cluster solution was obtained based on unique NS profiles, and divided patients into low NS (LNS; 42.6%), moderate NS (MNS; 25.7%) and high NS (HNS; 31.7%). We next examined associations of the three unveiled NS clusters to clinically relevant external validity variables. Results indicated that the identified subgroups significantly differed on state depression, trait depression, loneliness and social support.
Our sample of 202 chronic inpatients was divided into three distinct homogeneous profiles (LNS, MNS, HNS) based on the five NS domains (social withdrawal, emotional withdrawal, alogia, avolition, anhedonia) used for clustering. These findings expand and corroborate previous research using various cluster analytic approaches and carried out mainly in Western countries [
42]. While extensive research efforts have been devoted to dissecting the heterogeneity of schizophrenia using statistical subgrouping methods [
27,
95‐
97], only few empirical evidence sought to determine meaningful boundaries within the disease based on NS (e.g [
43‐
45].,). Using a similar procedure to ours, Paul et al. [
52] were the first to deconstruct heterogeneity in schizophrenia through the five NIMH NS domains (i.e., anhedonia, asociality, avolition, blunted affect, and alogia) in a sample of 220 outpatients meeting criteria for schizophrenia or schizoaffective disorder from the USA. Authors indicated a four-cluster solution as optimal, with patients being divided into two subgroups with either low or severe NS levels, and two other subgroups with moderate NS levels and either increased blunted affect or diminished avolition [
52]. Differences in identification of symptom clusters between this study and ours is likely, in part, due to cultural factors as well as symptom-reporting method (clinical interview vs. self-report). The largest number of our participants belonged to the LNS cluster (42.6%), which is consistent with recent previous findings by Paul et al. [
52], and further endorses observations reported in prior cluster analysis studies that there exists a cluster with low or transient NS [
22,
43‐
45].
The cluster analysis approach assumes that the three clusters of patients identified display more between-cluster that within-cluster variation [
98]. We thus had as a second objective to investigate whether our three groups differed meaningfully from one another in their associations with external validators. To this end, we compared their performance on demographic (age, gender, marital status, educational level) and clinical (diagnosis, duration of illness, duration of hospitalization, antipsychotics dose) factors, as well as other assessments of state depression, trait depression, stigma, insight, loneliness and social support. Post-hoc comparisons showed that depression (state and trait), loneliness and social support could accurately distinguish the schizophrenia subgroups. Additionally, individuals in the HNS cluster had longer duration of illness, longer duration of hospitalization, and were given higher dosages of antipsychotic medication compared to those in the other clusters, but these differences did not achieve the statistical significance. Consistent with our results, a previous study using latent class growth analysis to model changes in NS over a 12-month follow-up period found that depression predicted initially high NS in a large sample of patients with first episode psychosis [
99]. There is a phenomenological overlap between negative and depressive symptomatology, and a substantial lack of clarity on how to reliably assess them and how to validly distinguish between them [
100]. Depressive symptoms are (more often) defined by self-report criteria [
101]. Older clinician-rated measures (e.g., the Scale for the Assessment of Negative Symptoms [
102]) conceptualize NS as a single construct including multiple symptoms, and do not discriminate experiential NS that are commonly seen in depression (i.e. low motivation, anhedonia and withdrawal) from expressive symptoms (i.e. blunted affect and alogia) [
101]. Newer measures, such as the SNS used in the present study to assess NS, have demonstrated good divergent validity from depressive measures [
54,
58,
59]. In sum, these observations are further in agreement with prior findings that, despite the theoretical overlap in clinical presentation between NS and depression (e.g., social withdrawal, anhedonia, apathy, diminished emotional expression) often leading to a diagnostic dilemma between the two entities [
103,
104], they are distinct and separate symptom domains according to factor analysis studies [
105]. Furthermore, a variety of depressive symptoms may overlap with certain other features common to schizophrenia, including neuroleptic induced side effects [
106] or the negative effects of long-term hospitalization. Therefore, our findings should be considered tentative until confirmed by future studies involving outpatient populations and using clinical interview to assess both depression and NS.
The predicting effect of social support and loneliness is also consistent with prior research on NS, which suggests that the deficit schizophrenia group (exhibiting high NS levels) appear to suffer from a marked lack of interpersonal relatedness [
107], as well as great difficulties in social contact and interest [
108]. In accordance with our findings, a previous study showed that NS groups were not distinguishable by perceived level of internalized stigma [
108]. Nevertheless, it was somewhat unexpected to find no differences between subgroups regarding some external validators, including level of education, duration of illness, duration of hospitalization, antipsychotics dose, as previous studies have established such associations [
22,
52]. Therefore, additional future studies are necessary to elucidate the separation between clusters on these factors.
Clinical implications and research perspectives
Our cluster analysis proposes that subtypes of schizophrenia may exist with severity-based differences in underlying NS. The HNS subgroup had the greatest levels of trait/state depression and loneliness, and the lowest levels of perceived social support. Altogether, findings advance that schizophrenia encompasses qualitatively separate NS subgroups that differ in their psychopathological profiles. Heterogeneity in schizophrenia may echo a combination of homogeneous, non-arbitrary subgroups that, when taken into account, shed light on different etiological processes and guide efforts to develop more effective and more specific treatments based on group-level characteristics [
42]. Indeed, the identification of schizophrenia subgroups could assist in advancing evidence-based personalized medicine in the field of schizophrenia and related psychoses by selecting and applying treatment options appropriate for subtypes of patients with similar and unique features. Finally, the identification of depression, loneliness and social support as clinically relevant predictors offers promising avenues to develop clinical risk (and machine learning) prediction models [
42].
Study limitations
Despite its significant contribution to the international literature, this study has certain limitations that need to be acknowledged. First, our study included long-stay chronic inpatients, predominantly males (62.4%), with a long mean duration of illness (over 30 years) from Lebanon, which may affect the generalizability of our findings to outpatients, younger, female patients in early phases of illness, as well as those in other countries and cultures. Second, data was gathered at a single point in time, which prevented investigation of the stability of negative symptom profiles across phases of illness. Future longitudinal studies are required to address this point. Third, although self-evaluation of NS by the patients themselves has its advantages, it remains subjective and should be complemented by semi-structured interviews. Fourth, our external validation data did not include other important clinical variables (e.g., positive symptoms), neurocognitive symptoms, psychosocial functioning, quality of life, genetic and structural neuroimaging factors; resulting in limited information about qualitative differences across the three subgroups. More research should consider including additional external validators. Another important limitation to the present study is that it did not examine whether the links found are equivalent across all different forms of NS, such as primary vs. secondary NS, especially since these subgroups vary on some of the external validator variables including positive psychotic symptoms, substance use, social deprivation, and depression [
37,
109,
110]. Although previous studies investigating NS subgroups did neither differentiate primary from secondary symptoms, nor did they control for the major sources of secondary NS (e.g [
50,
52].,), we recommend that future research differentiate between these different forms of NS. In addition, our sample consists of long-stay hospitalized patients with an average duration of hospitalization of 28 years, which might have affected our findings, as residence in an institution is likely to contribute to secondary NS [
111]. Prolonged hospital stay could also significantly affect perceived social support and loneliness experiences [
112]. Therefore, further studies need to replicate our findings in outpatients.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.